There are currently thousands of Jamaican citizens that lack access to basic health care. In order to improve the health-care system, I collect and analyze data from two clinics in remote locations of the island. This report analyzes data collected from Clarendon Parish, Jamaica. In order to create a descriptive analysis, I use SAS^® Studio 9.4. A few of the procedures I use include: PROC IMPORT, PROC MEANS, PROC FREQ, and PROC GCHART. After conducting the aforementioned procedures, I am able to produce a descriptive analysis of the health issues plaguing the island.

Read the paper (PDF) | View the e-poster or slides (PDF)

Although rapid development of information technologies in the past decade has provided forecasters with both huge amounts of data and massive computing capabilities, these advancements do not necessarily translate into better forecasts. Different industries and products have unique demand patterns. There is not yet a one-size-fits-all forecasting model or technique. A good forecasting model must be tailored to the data in order to capture the salient features and satisfy the business needs. This paper presents a multistage modeling strategy for demand forecasting. The proposed strategy, which has no restrictions on the forecasting techniques or models, provides a general framework for building a forecasting system in multiple stages.

Although the statistical foundations of predictive analytics have large overlaps, the business objectives, data availability, and regulations are different across the property and casualty insurance, life insurance, banking, pharmaceutical, and genetics industries. A common process in property and casualty insurance companies with large data sets is introduced, including data acquisition, data preparation, variable creation, variable selection, model building (also known as fitting), model validation, and model testing. Variable selection and model validation stages are described in more detail. Some successful models in the insurance companies are introduced. Base SAS^®, SAS^® Enterprise Guide^®, and SAS^® Enterprise Miner™ are presented as the main tools for this process.

This paper aims to show a SAS^® macro for generating random numbers of skew-normal and skew-t distributions, as well as the quantiles of these distributions. The results are similar to those generated by the sn package of R software.

Read the paper (PDF) | Download the data file (ZIP) | View the e-poster or slides (PDF)

This work presents a macro for generating a ternary graph that can be used to solve a problem in the ceramics industry. The ceramics industry uses mixtures of clay types to generate a product with good properties that can be assessed according to breakdown pressure after incineration, the porosity, and water absorption. Beyond these properties, the industry is concerned with managing geological reserves of each type of clay. Thus, it is important to seek alternative compositions that present properties similar to the default mixture. This can be done by analyzing the surface response of these properties according to the clay composition, which is easily done in a ternary graph. SAS^® documentation does not describe how to adjust an analysis grid in nonrectangular forms on graphs. A macro triaxial, however, can generate ternary graphical analysis by creating a special grid, using an annotated database, and making linear transformations of the original data. The GCONTOUR procedure is used to generate three-dimensional analysis.

Read the paper (PDF) | Watch the recording

Geographically Weighted Negative Binomial Regression (GWNBR) was developed by Silva and Rodrigues (2014). It is a generalization of the Geographically Weighted Poisson Regression (GWPR) proposed by Nakaya and others (2005) and of the Poisson and negative binomial regressions. This paper shows a SAS^® macro to estimate the GWNBR model encoded in SAS/IML^® software and shows how to use the SAS procedure GMAP to draw the maps.

Read the paper (PDF) | Download the data file (ZIP)

It is common for hundreds or even thousands of clinical endpoints to be collected from individual subjects, but events from the majority of clinical endpoints are rare. The challenge of analyzing high dimensional sparse data is in balancing analytical consideration for statistical inference and clinical interpretation with precise meaningful outcomes of interest at intra-categorical and inter-categorical levels. Lumping or grouping similar rare events into a composite category has the statistical advantage of increasing testing power, reducing multiplicity size, and avoiding competing risk problems. However, too much or inappropriate lumping would jeopardize the clinical meaning of interest and external validity, whereas splitting or keeping each individual event at its basic clinical meaningful category can overcome the drawbacks of lumping. This practice might create analytical issues of increasing type II errors, multiplicity size, competing risks, and having a large proportion of endpoints with rare events. It seems that lumping and splitting are diametrically opposite approaches, but in fact, they are complementary. Both are essential for high dimensional data analysis. This paper describes the steps required for the lumping and splitting analysis and presents SAS^® code that can be used to implement each step.

Longitudinal and repeated measures data are seen in nearly all fields of analysis. Examples of this data include weekly lab test results of patients or performance of test score by children from the same class. Statistics students and analysts alike might be overwhelmed when it comes to repeated measures or longitudinal data analyses. They might try to educate themselves by diving into text books or taking semester-long or intensive weekend courses, resulting in even more confusion. Some might try to ignore the repeated nature of data and take short cuts such as analyzing all data as independent observations or analyzing summary statistics such as averages or changes from first to last points and ignoring all the data in-between. This hands-on presentation introduces longitudinal and repeated measures analyses without heavy emphasis on theory. Students in the workshop will have the opportunity to get hands-on experience graphing longitudinal and repeated measures data. They will learn how to approach these analyses with tools like PROC MIXED and PROC GENMOD. Emphasis will be on continuous outcomes, but categorical outcomes will briefly be covered.

Government funding is a highly sought-after financing medium by entrepreneurs and researchers aspiring to make a breakthrough in the market and compete with larger organizations. The funding is channeled via federal agencies that seek out promising research and products that improve the extant work. This study analyzes the project abstracts that earned the government funding through a text analytic approach over the past three and a half decades in the fields of Defense and Health and Human Services. This helps us understand the factors that might be responsible for awards made to a project. We collected the data of 100,279 records from the Small Business Innovation and Research (SBIR) website (https://www.sbir.gov). Initially, we analyze the trends in government funding over research by the small businesses. Then we perform text mining on the 55,791 abstracts of the projects that are funded by the Department of Defense and the Department of Health. From the text mining, we portray the key topics and patterns related to the past research and determine the changes in their trends over time. Through our study, we highlight the research trends to enable organizations to shape their business strategy and make better investments in the future research.

In the Collection Direction of a well-recognized Colombian financial institution, there was no methodology that provided an optimal number of collection agents to improve the collection task and make possible that more customers be compromised with their minimum monthly payment of their debt. The objective of this paper is to apply the Data Envelopment Analysis (DEA) Optimization Methodology to determine the optimal number of agents to maximize the monthly collection in the bank. We show that the results can have a positive impact to the credit portfolio behavior and reduce the collection management cost. DEA optimization methodology has been successfully used in various fields to solve multi-criteria optimization problems, but it is not commonly used in the financial sector mostly because this methodology requires specialized software, such as SAS^® Enterprise Guide^®. In this paper, we present the PROC OPTMODEL and we show how to formulate the optimization problem, program the SAS^® Code, and how to process adequately the available data.

Read the paper (PDF) | Download the data file (ZIP)

Currently Colpatria, as a part of Scotiabank in Colombia, has several methodologies that enable us to have a vision of the customer from a risk perspective. However, the current trend in the financial sector is to have a global vision that involves aspects of risk as well as of profitability and utility. As a part of the business strategies to develop cross-sell and customer profitability under conditions of risk needs, it's necessary to create a customer value index to score the customer according to different groups of business key variables that permit us to describe the profitability and risk of each customer. In order to generate the Index of Customer Value, we propose to construct a synthetic index using principal component analysis and multiple factorial analysis.

In the regulatory world of patient safety and pharmacovigilance, whether it's during clinical trials or post-market surveillance, SAEs that affect participants must be collected, and if certain criteria are met, reported to the FDA and other regulatory authorities. SAEs are often entered into multiple databases by various users, resulting in possible data discrepancies and quality loss. Efforts have been made to reconcile the SAE data between databases, but there is no industrial standard regarding the methodology or tool employed for this task. Some organizations still reconcile the data manually, with visual inspections and vocal verification. Not only is this laborious and error-prone, it becomes prohibitive when the data reach hundreds of records. We devised an efficient algorithm using SAS^® to compare two data sources automatically. Our algorithm identifies matched, discrepant, and unpaired SAE records. Additionally, it employs a user-supplied list of synonyms to find non-identical but relevant matches. First, two data sources are combined and sorted by key fields such as Subject ID , Onset Date , Stop Date , and Event Term . Record counts and Levenshtein edit distances are calculated within certain groups to assist with sorting and matching. This combined record list is then fed into a DATA step to decide whether a record is paired or unpaired. For an unpaired record, a stub record with all fields set as ? is generated as a matching placeholder. Each record is written to one of two data sets. Later, the data sets are tagged and pulled into a comparison logic using hash objects, which enable field-by-field comparison and display discrepancies in clean format for each field. Identical fields or columns are cleared or removed for clarity. The result is a streamlined and user-friendly process that allows for fast and easy SAE reconciliation.

This paper presents supervised and unsupervised pattern recognition techniques that use Base SAS^® and SAS^® Enterprise Miner™ software. A simple preprocessing technique creates many small image patches from larger images. These patches encourage the learned patterns to have local scale, which follows well-known statistical properties of natural images. In addition, these patches reduce the number of features that are required to represent an image and can decrease the training time that algorithms need in order to learn from the images. If a training label is available, a classifier is trained to identify patches of interest. In the unsupervised case, a stacked autoencoder network is used to generate a dictionary of representative patches, which can be used to locate areas of interest in new images. This technique can be applied to pattern recognition problems in general, and this paper presents examples from the oil and gas industry and from a solar power forecasting application.

The world's capacity to store and analyze data has increased in ways that would have been inconceivable just a couple of years ago. Due to this development, large-scale data are collected by governments. Until recently, this was for purely administrative purposes. This study used comprehensive data files on education. The purpose of this study was to examine compulsory courses for a bachelor's degree in the economics program at the University of Copenhagen. The difficulty and use of the grading scale was compared across the courses by using the new IRT procedure, which was introduced in SAS/STAT^® 13.1. Further, the latent ability traits that were estimated for all students in the sample by PROC IRT are used as predictors in a logistic regression model. The hypothesis of interest is that students who have a lower ability trait will have a greater probability of dropping out of the university program compared to successful students. Administrative data from one cohort of students in the economics program at the University of Copenhagen was used (n=236). Three unidimensional Item Response Theory models, two dichotomous and one polytomous, were introduced. It turns out that the polytomous Graded Response model does the best job of fitting data. The findings suggest that in order to receive the highest possible grade, the highest level of student ability is needed for the course exam in the first-year course Descriptive Economics A. In contrast, the third-year course Econometrics C is the easiest course in which to receive a top grade. In addition, this study found that as estimated student ability decreases, the probability of a student dropping out of the bachelor's degree program increases drastically. However, contrary to expectations, some students with high ability levels also end up dropping out.

Movie reviews provide crucial information and act as an important factor when deciding whether to spend money on seeing a film in the theater. Each review reflects an individual's take on the movie and there are often contrasting reviews for the same movie. Going through each review may create confusion in the mind of a reader. Analyzing all the movie reviews and generating a quick summary that describes the performance, direction, and screenplay, among other aspects is helpful to the readers in understanding the sentiments of movie reviewers and in deciding if they would like to watch the movie in the theaters. In this paper, we demonstrate how the SAS^® Enterprise Miner™ nodes enable us to generate a quick summary of the terms and their relationship with each other, which describes the various aspects of a movie. The Text Cluster and Text Topic nodes are used to generate groups with similar subjects such as genres of movies, acting, and music. We use SAS^® Sentiment Analysis Studio to build models that help us classify 10,000 reviews where each review is a separate document and the reviews are equally split into good or bad. The Smoothed Relative Frequency and Chi-square model is found to be the best model with overall precision of 78.37%. As soon as the latest reviews are out, such analysis can be performed to help viewers quickly grasp the sentiments of the reviews and decide if the movie is worth watching.

When dealing with non-normal categorical response variables, logistic regression is the robust method to use for modeling the relationship between categorical outcomes and different predictors without assuming a linear relationship between them. Within such models, the categorical outcome might be binary, multinomial, or ordinal, and predictors might be continuous or categorical. Another complexity that might be added to such studies is when data is longitudinal, such as when outcomes are collected at multiple follow-up times. Learning about modeling such data within any statistical method is beneficial because it enables researchers to look at changes over time. This study looks at several methods of modeling binary and categorical response variables within regression models by using real-world data. Starting with the simplest case of binary outcomes through ordinal outcomes, this study looks at different modeling options within SAS^® and includes longitudinal cases for each model. To assess binary outcomes, the current study models binary data in the absence and presence of correlated observations under regular logistic regression and mixed logistic regression. To assess multinomial outcomes, the current study uses multinomial logistic regression. When responses are ordered, using ordinal logistic regression is required as it allows for interpretations based on inherent rankings. Different logit functions for this model include the cumulative logit, adjacent-category logit, and continuation ratio logit. Each of these models is also considered for longitudinal (panel) data using methods such as mixed models and Generalized Estimating Equations (GEE). The final consideration, which cannot be addressed by GEE, is the conditional logit to examine bias due to omitted explanatory variables at the cluster level. Different procedures for the aforementioned within SAS^® 9.4 are explored and their strengths and limitations are specified for applied researchers finding simil ar data characteristics. These procedures include PROC LOGISTIC, PROC GLIMMIX, PROC GENMOD, PROC NLMIXED, and PROC PHREG.

Breast cancer is the second leading cause of cancer deaths among women in the United States. Although mortality rates have been decreasing over the past decade, it is important to continue to make advances in diagnostic procedures as early detection vastly improves chances for survival. The goal of this study is to accurately predict the presence of a malignant tumor using data from fine needle aspiration (FNA) with visual interpretation. Compared with other methods of diagnosis, FNA displays the highest likelihood for improvement in sensitivity. Furthermore, this study aims to identify the variables most closely associated with accurate outcome prediction. The study utilizes the Wisconsin Breast Cancer data available within the UCI Machine Learning Repository. The data set contains 699 clinical case samples (65.52% benign and 34.48% malignant) assessing the nuclear features of fine needle aspirates taken from patients' breasts. The study analyzes a variety of traditional and modern models, including: logistic regression, decision tree, neural network, support vector machine, gradient boosting, and random forest. Prior to model building, the weights of evidence (WOE) approach was used to account for the high dimensionality of the categorical variables after which variable selection methods were employed. Ultimately, the gradient boosting model utilizing a principal component variable reduction method was selected as the best prediction model with a 2.4% misclassification rate, 100% sensitivity, 0.963 Kolmogorov-Smirnov statistic, 0.985 Gini coefficient, and 0.992 ROC index for the validation data. Additionally, the uniformity of cell shape and size, bare nuclei, and bland chromatin were consistently identified as the most important FNA characteristics across variable selection methods. These results suggest that future research should attempt to refine the techniques used to determine these specific model inputs. Greater accuracy in characterizing the FNA attributes will allow researchers to develop more promising models for early detection.

Read the paper (PDF) | View the e-poster or slides (PDF)

Using smart clothing with wearable medical sensors integrated to keep track of human health is now attracting many researchers. However, body movement caused by daily human activities inserts artificial noise into physiological data signals, which affects the output of a health monitoring/alert system. To overcome this problem, recognizing human activities, determining relationship between activities and physiological signals, and removing noise from the collected signals are essential steps. This paper focuses on the first step, which is human activity recognition. Our research shows that no other study used SAS^® for classifying human activities. For this study, two data sets were collected from an open repository. Both data sets have 561 input variables and one nominal target variable with four levels. Principal component analysis along with other variable reduction and selection techniques were applied to reduce dimensionality in the input space. Several modeling techniques with different optimization parameters were used to classify human activity. The gradient boosting model was selected as the best model based on a test misclassification rate of 0.1233. That is, 87.67% of total events were classified correctly.

Read the paper (PDF) | View the e-poster or slides (PDF)

Massive Open Online Courses (MOOC) have attracted increasing attention in educational data mining research areas. MOOC platforms provide free higher education courses to Internet users worldwide. However, MOOCs have high enrollment but notoriously low completion rates. The goal of this study is apply frequentist and Bayesian logistic regression to investigate whether and how students' engagement, intentions, education levels, and other demographics are conducive to course completion in MOOC platforms. The original data used in this study came from an online eight-week course titled Big Data in Education, taught within the Coursera platform (MOOC) by Teachers College, Columbia University. The data sets for analysis were created from three different sources--clickstream data, a pre-course survey, and homework assignment files. The SAS system provides multiple procedures to perform logistic regression, with each procedure having different features and functions. In this study, we apply two approaches--frequentist and Bayesian logistic regression, to MOOC data. PROC LOGISTIC is used for the frequentist approach, and PROC GENMOD is used for Bayesian analysis. The results obtained from the two approaches are compared. All the statistical analyses are conducted in SAS^® 9.3. Our preliminary results show that MOOC students with higher course engagement and higher motivation are more likely to complete the MOOC course.

Read the paper (PDF) | View the e-poster or slides (PDF)

Arrays and DO loops have been used for years by SAS^® programmers who work with diagnosis fields in health-care data, and the opportunity to use these techniques has only grown since the launch of the Affordable Care Act (ACA) in the United States. Users new to SAS or to the health-care field may find an overview of existing (as well as new) applications helpful. Risk-adjustment software, including the publicly available Health and Human Services (HHS) risk software that uses SAS and was released as part of the ACA implementation, is one example of code that is significantly improved by the use of arrays. Similar projects might include evaluations of diagnostic persistence, comparisons of diagnostic frequency or intensity between providers, and checks for unusual clusters of diagnosed conditions. This session reviews examples suitable for intermediate SAS users, including the application of two-dimensional arrays to diagnosis fields.

In light of past indiscretions by the police departments leading to riots in major U.S. cities, it is important to assess factors leading to the arrest of a citizen. The police department should understand the arrest situation before they can make any decision and diffuse any possibility of a riot or similar incidents. Many such incidents in the past are a result of the police department failing to make right decisions in the case of emergencies. The primary objective of this study is to understand the key factors that impact arrest of people in New York City and understanding various crimes that have taken place in the region in the last two years. The study explores different regions of New York City where the crimes have taken place and also the timing of incidents leading to them. The data set consists of 111 variables and 273,430 observations from the year 2013 to 2014, with a binary target variable arrest made . The percentage of Yes and No incidents of arrest made are 6% and 94% respectively. This study analyzes the reasons for arrest, which are suspicion, timing of frisk activities, identifying threats, basis of search, whether arrest required a physical intervention, location of the arrest, and whether the frisk officer needs any support. Decision tree, regression, and neural network models were built, and decision tree turned out to be the best model with a validation misclassification rate of 6.47% and model sensitivity of 73.17%. The results from the decision tree model showed that suspicion count, search basis count, summons leading to violence, ethnicity, and use of force are some of the important factors influencing arrest of a person.

Bayesian inference for complex hierarchical models with smoothing splines is typically intractable, requiring approximate inference methods for use in practice. Markov Chain Monte Carlo (MCMC) is the standard method for generating samples from the posterior distribution. However, for large or complex models, MCMC can be computationally intensive, or even infeasible. Mean Field Variational Bayes (MFVB) is a fast deterministic alternative to MCMC. It provides an approximating distribution that has minimum Kullback-Leibler distance to the posterior. Unlike MCMC, MFVB efficiently scales to arbitrarily large and complex models. We derive MFVB algorithms for Gaussian semiparametric multilevel models and implement them in SAS/IML^® software. To improve speed and memory efficiency, we use block decomposition to streamline the estimation of the large sparse covariance matrix. Through a series of simulations and real data examples, we demonstrate that the inference obtained from MFVB is comparable to that of PROC MCMC. We also provide practical demonstrations of how to estimate additional posterior quantities of interest from MFVB either directly or via Monte Carlo simulation.

Read the paper (PDF) | Download the data file (ZIP)

Building representative machine learning models that generalize well on future data requires careful consideration both of the data at hand and of assumptions about the various available training algorithms. Data are rarely in an ideal form that enables algorithms to train effectively. Some algorithms are designed to account for important considerations such as variable selection and handling of missing values, whereas other algorithms require additional preprocessing of the data or appropriate tweaking of the algorithm options. Ultimate evaluation of a model's quality requires appropriate selection and interpretation of an assessment criterion that is meaningful for the given problem. This paper discusses the most common issues faced by machine learning practitioners and provides guidance for using these powerful algorithms to build effective models.

Higher education is taking on big data initiatives to help improve student outcomes and operational efficiencies. This paper shows how one institution went form White Paper to action. Challenges such as developing a common definition of big data, access to certain types of data, and bringing together institutional groups who had little previous contact are highlighted. The stepwise, collaborative approach taken is highlighted. How the data was eventually brought together and analyzed to generate actionable results is also covered. Learn how to make big data a part of your analytics toolbox.

The surge of data and data sources in marketing has created an analytical bottleneck in most organizations. Analytics departments have been pushed into a difficult decision: either purchase black-box analytical tools to generate efficiencies or hire more analysts, modelers, and data scientists. Knowledge gaps stemming from restrictions in black-box tools or from backlogs in the work of analytical teams have resulted in lost business opportunities. Existing big data analytics tools respond well when dealing with large record counts and small variable counts, but they fall short in bringing efficiencies when dealing with wide data. This paper discusses the importance of an agile modeling engine designed to deliver productivity, irrespective of the size of the data or the complexity of the modeling approach.

Read the paper (PDF) | Watch the recording

Nowadays, the recommender system is a popular tool for online retailer businesses to predict customers' next-product-to-buy (NPTB). Based on statistical techniques and the information collected by the retailer, an efficient recommender system can suggest a meaningful NPTB to customers. A useful suggestion can reduce the customer's searching time for a wanted product and improve the buying experience, thus increasing the chance of cross-selling for online retailers and helping them build customer loyalty. Within a recommender system, the combination of advanced statistical techniques with available information (such as customer profiles, product attributes, and popular products) is the key element in using the retailer's database to produce a useful suggestion of an NPTB for customers. This paper illustrates how to create a recommender system with the SAS^® RECOMMEND procedure for online business. Using the recommender system, we can produce predictions, compare the performance of different predictive models (such as decision trees or multinomial discrete-choice models), and make business-oriented recommendations from the analysis.

Read the paper (PDF) | Download the data file (ZIP)

Even though marketing is inevitable in every business, every year the marketing budget is limited and prudent fund allocations are required to optimize marketing investment. In many businesses, the marketing fund is allocated based on the marketing manager's experience, departmental budget allocation rules, and sometimes 'gut feelings' of business leaders. Those traditional ways of budget allocation yield suboptimal results and in many cases lead to money being wasted on irrelevant marketing efforts. Marketing mixed models can be used to understand the effects of marketing activities and identify the key marketing efforts that drive the most sales among a group of competing marketing activities. The results can be used in marketing budget allocation to take out the guesswork that typically goes into the budget allocation. In this paper, we illustrate practical methods for developing and implementing marketing mixed modeling using SAS^® procedures. Real-life challenges of marketing mixed model development and execution are discussed, and several recommendations are provided to overcome some of those challenges.

We set out to model two of the leading causes of fatal automobile accidents in America - drunk driving and speeding. We created a decision tree for each of those causes that uses situational conditions such as time of day and type of road to classify historical accidents. Using this model a law enforcement or town official can decide if a speed trap or DUI checkpoint can be the most effective in a particular situation. This proof of concept can easily be applied to specific jurisdictions for customized solutions.

Thanks to advances in technologies that make data more readily available, sports analytics is an increasingly popular topic. A majority of sports analyses use advanced statistics and metrics to achieve their goal, whether it be prediction or explanation. Few studies include public opinion data. Last year's highly anticipated NCAA College Football Championship game between Ohio State and Oregon broke ESPN and cable television records with an astounding 33.4 million viewers. Given the popularity of college football, especially now with the inclusion of the new playoff system, people seem to be paying more attention than ever to the game. ESPN provides fans with College Pick'em, which gives them a way to compete with their friends and colleagues on a weekly basis, for free, to see who can correctly pick the winners of college football games. Each week, 10 close matchups are selected, and users must select which team they think will win the game and rank those picks on a scale of 1 (lowest) to 10 (highest), according to their confidence level. For each team, the percentage of users who picked that team and the national average confidence are shown. Ideally, one could use these variables in conjunction with other information to enhance one's own predictions. The analysis described in this session explores the relationship between public opinion data from College Pick'em and the corresponding game outcomes by using visualizations and statistical models implemented by various SAS^® products.

Fire department facilities in Belgium are in desperate need of renovation due to relentless budget cuts during the last decade. But time has changed and the current location is troubled by city center traffic jams or narrow streets. In other words, their current location is not ideal. The approach for our project was to find the best possible locations for new facilities, based on historical interventions and live traffic information for fire trucks. The goal was not to give just one answer, but to provide instead SAS^® Visual Analytics dashboards that allowed policy makers to play with some of the deciding factors and view the impact of cutting off or adding facilities on the estimated intervention time. For each option the optimal locations are given. SAS^® software offers many great products but the true power of SAS as a data mining tool is in using the full SAS suite. SAS^® Data Integration Studio was used for its parallel processing capabilities in extracting and parsing the traveling time via an online API to MapQuest. SAS/OR^® was used for the optimization of the model using a Lagrangian multiplier. SAS^® Enterprise Guide^® was used to combine and integrate the data preparation. SAS Visual Analytics was used to visualize the results and to serve as an interface to execute the model via a stored process.

One of the major diseases that records a high number of readmissions is bacterial pneumonia in Medicare patients. This study aims at comparing and contrasting Northeast and South regions of the United States based on the factors causing the 30-day readmissions. The study also identifies some of the ICD-9 medical procedures associated with those readmissions. Using SAS^® Enterprise Guide^® 7.1 and SAS^® Enterprise Miner™ 14.1, this study analyzes patient and hospital demographics of the Northeast and South regions of United States using the Cerner Health Facts Database. Further, the study suggests some preventive measures to reduce readmissions. The 30-day readmissions are computed based on admission and discharge dates from 2003 until 2013. Using clustering, various hospitals, along with discharge disposition levels (where a patient is sent after discharge), are grouped. In both regions, the patients who are discharged to home have shown significantly lower chances of readmission. The odds ratio = 0.562 for the Northeast region and for the South region; all other disposition levels have significantly high odds ( > 1), compared to discharged to home. For the South region, females have around 53 percent more possibility of readmission compared to males (odds ratio = 1.535). Also some of the hospital groups have higher readmission cases. The association analysis using the Market Basket node finds catheterization procedures to be highly significant (Lift = 1.25 for the Northeast and 3.84 for the South; Confidence = 52.94% for the Northeast and 85.71% for the South) in bacterial pneumonia readmissions. By research it was found that during these procedures, patients are highly susceptible to acquiring Methicillin-resistant Staphylococcus aureus (MRSA) bacteria, which causes Methicillin-susceptible pneumonia. Providing timely follow up for the patients operated with these procedures might possibly limit readmissions. These patients might also be discharge d to home under medical supervision, as such patients had shown significantly lower chances of readmission.

Competing-risks analyses are methods for analyzing the time to a terminal event (such as death or failure) and its cause or type. The cumulative incidence function CIF(j, t) is the probability of death by time t from cause j. New options in the LIFETEST procedure provide for nonparametric estimation of the CIF from event times and their associated causes, allowing for right-censoring when the event and its cause are not observed. Cause-specific hazard functions that are derived from the CIFs are the analogs of the hazard function when only a single cause is present. Death by one cause precludes occurrence of death by any other cause, because an individual can die only once. Incorporating explanatory variables in hazard functions provides an approach to assessing their impact on overall survival and on the CIF. This semiparametric approach can be analyzed in the PHREG procedure. The Fine-Gray model defines a subdistribution hazard function that has an expanded risk set, which consists of individuals at risk of the event by any cause at time t, together with those who experienced the event before t from any cause other than the cause of interest j. Finally, with additional assumptions a full parametric analysis is also feasible. We illustrate the application of these methods with empirical data sets.

Read the paper (PDF) | Watch the recording

Abstract quasi-complete separation is a commonly detected issue in logit/probit models. Quasi-complete separation occurs when a dependent variable separates an independent variable or a combination of several independent variables to a certain degree. In other words, levels in a categorical variable or values in a numeric variable are separated by groups of discrete outcome variables. Most of the time, this happens in categorical independent variable(s). Quasi-complete separation can cause convergence failures in logistic regression, which consequently result in a large coefficient estimate and standard errors inflation. Therefore, it can potentially yield biased results. This paper provides comprehensive reviews with various approaches for correcting quasi-complete separation in binary logistic regression model and presents hands-on examples from the healthcare insurance industry. First, it introduces the concept of quasi-complete separation and how to diagnosis the issue. Then, it provides step-by-step guidelines for fixing the problem, from a straightforward data configuration approach to a complicated statistical modeling approach such as the EXACT method, the FIRTH method, and so on.

Credit card profitability prediction is a complex problem because of the variety of card holders' behavior patterns and the different sources of interest and transactional income. Each consumer account can move to a number of states such as inactive, transactor, revolver, delinquent, or defaulted. This paper i) describes an approach to credit card account-level profitability estimation based on the multistate and multistage conditional probabilities models and different types of income estimation, and ii) compares methods for the most efficient and accurate estimation. We use application, behavioral, card state dynamics, and macroeconomic characteristics, and their combinations as predictors. We use different types of logistic regression such as multinomial logistic regression, ordered logistic regression, and multistage conditional binary logistic regression with the LOGISTIC procedure for states transition probability estimation. The state transition probabilities are used as weights for interest rate and non-interest income models (which one is applied depends on the account state). Thus, the scoring model is split according to the customer behavior segment and the source of generated income. The total income consists of interest and non-interest income. Interest income is estimated with the credit limit utilization rate models. We test and compare five proportion models with the NLMIXED, LOGISTIC, and REG procedures in SAS/STAT^® software. Non-interest income depends on the probability of being in a particular state, the two-stage model of conditional probability to make a point-of-sales transaction (POS) or cash withdrawal (ATM), and the amount of income generated by this transaction. We use the LOGISTIC procedure for conditional probability prediction and the GLIMMIX and PANEL procedures for direct amount estimation with pooled and random-effect panel data. The validation results confirm that traditional techniques can be effectively applied to complex tasks with many para meters and multilevel business logic. The model is used in credit limit management, risk prediction, and client behavior analytics.

The recent advances in regulatory stress testing, including stress testing regulated by Comprehensive Capital Analysis and Review (CCAR) in the US, the Prudential Regulation Authority (PRA) in the UK, and the European Banking Authority in the EU, as well as the new international accounting requirement known as IFRS 9 (International Financial Reporting Standard), all pose new challenges to credit risk modeling. The increasing sophistication of the models that are supposed to cover all the material risks in the underlying assets in various economic scenarios makes models harder to implement. Banks are spending a lot of resources on the model implementation but are still facing issues due to long time to deployment and disconnection between the model development and implementation teams. Models are also required at a more granular level, in many cases, down to the trade and account levels. Efficient model execution becomes valuable for banks to get timely response to the analysis requests. At the same time, models are subject to more stringent internal and external scrutiny. This paper introduces a suite of risk modeling solutions from credit risk modeling leader SAS^® to help banks overcome these new challenges and be competent to meet the regulatory requirements.

Customer retention is a primary concern for businesses that rely on a subscription revenue model. It is common for marketers of subscription-based offerings to develop predictive models that are aimed at identifying subscribers who have the highest risk of attrition. With these likely unsubscribes identified, marketers then attempt to forestall customer termination by using a variety of retention enhancement tactics, which might include free offers, customer training, satisfaction surveys, or other measures. Although customer retention is always a worthy pursuit, it is often expensive to retain subscribers. In many cases, associated retention programs simply prove unprofitable over time because the overall cost of such programs frequently exceeds the lifetime value of the cohort of unsubscribed customers. Generally, it is more profitable to focus resources on identifying and marketing to a targeted prospective customer. When the target marketing strategy focuses on identifying prospects who are most likely to subscribe over the long term, the need for special retention marketing efforts decreases sharply. This paper describes results of an analytically driven targeting approach that is aimed at inviting new customers to a milk and grocery home-delivery service, with the promise of attracting only those prospects who are expected to exhibit high long-term retention rates.

Customer lifetime value (LTV) estimation involves two parts: the survival probabilities and profit margins. This article describes the estimation of those probabilities using discrete-time logistic hazard models. The estimation of profit margins is based on linear regression. In the scenario, when outliers are present among margins, we suggest applying robust regression with PROC ROBUSTREG.

Read the paper (PDF) | View the e-poster or slides (PDF)

Introduction: Cycling is on the rise in many urban areas across the United States. The number of cyclist fatalities is also increasing, by 19% in the last 3 years. With the broad-ranging personal and public health benefits of cycling, it is important to understand factors that are associated with these traffic-related deaths. There are more distracted drivers on the road than ever before, but the question remains of the extent that these drivers are affecting cycling fatality rates. Methods: This paper uses the Fatality Analysis Reporting System (FARS) data to examine factors related to cyclist death when the drivers are distracted. We use a novel machine learning approach, adaptive LASSO, to determine the relevant features and estimate their effect. Results: If a cyclist makes an improper action at or just before the time of the crash, the likelihood of the driver of the vehicle being distracted decreases. At the same time, if the driver is speeding or has failed to obey a traffic sign and fatally hits a cyclist, the likelihood of them also being distracted increases. Being distracted is related to other risky driving practices when cyclists are fatally injured. Environmental factors such as weather and road condition did not impact the likelihood that a driver was distracted when a cyclist fatality occurred.

Debt collection! The two words can trigger multiple images in one's mind--mostly harsh. However, let's try to think positively for a moment. In 2013, over $55 billion of debt were past due in the United States. What if all of these debts were left as is and the fate of credit issuers in the hands of good will payments made by defaulters? Well, this is not the most sustainable model, to say the least. In this situation, debt collection comes in as a tool that is employed at multiple levels of recovery to keep the credit flowing. Ranging from in-house to third-party to individual collection efforts, this industry is huge and plays an important role in keeping the engine of commerce running. In the recent past, with financial markets recovering and banks selling fewer charged-off accounts and at higher prices, debt collection has increasingly become a game of efficient operations backed by solid analytics. This paper takes you into the back alleys of all the data that is in there and gives an overview of some ways modeling can be used to impact the collection strategy and outcome. SAS^® tools such as SAS^® Enterprise Miner™ and SAS^® Enterprise Guide^® are extensively used for both data manipulation and modeling. Decision trees are given more focus to understand what factors make the most impact. Along the way, this paper also gives an idea of how analytics teams today are slowly trying to get the buy-in from other stakeholders in any company, which surprisingly is one of the most challenging aspects of our jobs.

Read the paper (PDF) | View the e-poster or slides (PDF)

Phishing is the attempt of a malicious entity to acquire personal, financial, or otherwise sensitive information such as user names and passwords from recipients through the transmission of seemingly legitimate emails. By quickly alerting recipients of known phishing attacks, an organization can reduce the likelihood that a user will succumb to the request and unknowingly provide sensitive information to attackers. Methods to detect phishing attacks typically require the body of each email to be analyzed. However, most academic institutions do not have the resources to scan individual emails as they are received, nor do they wish to retain and analyze message body data. Many institutions simply rely on the education and participation of recipients within their network. Recipients are encouraged to alert information security (IS) personnel of potential attacks as they are delivered to their mailboxes. This paper explores a novel and more automated approach that uses SAS^® to examine email header and transmission data to determine likely phishing attempts that can be further analyzed by IS personnel. Previously a collection of 2,703 emails from an external filtering appliance were examined with moderate success. This paper focuses on the gains from analyzing an additional 50,000 emails, with the inclusion of an additional 30 known attacks. Real-time email traffic is exported from Splunk Enterprise into SAS for analysis. The resulting model aids in determining the effectiveness of alerting IS personnel to potential phishing attempts faster than a user simply forwarding a suspicious email to IS personnel.

This paper presents an application based on predictive analytics and feature-extraction techniques to develop the alternative method for diagnosis of obstructive sleep apnea (OSA). Our method reduces the time and cost associated with the gold standard or polysomnography (PSG), which is operated manually, by automatically determining the OSA's severity of a patient via classification models using the time series from a one-lead electrocardiogram (ECG). The data is from Dr. Thomas Penzel of Philipps-University, Germany, and can be downloaded at www.physionet.org. The selected data consists of 10 recordings (7 OSAs, and 3 controls) of ECG collected overnight, and non-overlapping-minute-by-minute OSA episode annotations (apnea and non-apnea states). This accounts for a total of 4,998 events (2,532 non-apnea and 2,466 apnea minutes). This paper highlights the nonlinear decomposition technique, wavelet analysis (WA) in SAS/IML^® software, to maximize the information of OSA symptoms from ECG, resulting in useful predictor signals. Then, the spectral and cross-spectral analyses via PROC SPECTRA are used to quantify important patterns of those signals to numbers (features), namely power spectral density (PSD), cross power spectral density (CPSD), and coherency, such that the machine learning techniques in SAS^® Enterprise Miner™, can differentiate OSA states. To eliminate variations such as body build, age, gender, and health condition, we normalize each feature by the feature of its original signal (that is, ratio of PSD of ECGs WA by PSD of ECG). Moreover, because different OSA symptoms occur at different times, we account for this by taking features from adjacency minutes into analysis, and select only important ones using a decision tree model. The best classification result in the validation data (70:30) obtained from the Random Forest model is 96.83% accuracy, 96.39% sensitivity, and 97.26% specificity. The results suggest our method is well comparable to the gold standard.

While cardiac revascularization procedures like cardiac catheterization, percutaneous transluminal angioplasty, and cardiac artery bypass surgery have become standard practices in restorative cardiology, the practice is not evenly prescribed or subscribed to. We analyzed Florida hospital discharge records for the period 1992 to 2010 to determine the odds of receipt of any of these procedures by Hispanics and non-Hispanic Whites. Covariates (potential confounders) were age, insurance type, gender, and year of discharge. Additional covariates considered included comorbidities such as hypertension, diabetes, obesity, and depression. The results indicated that even after adjusting for covariates, Hispanics in Florida during the time period 1992 to 2010 were consistently less likely to receive these procedures than their White counterparts. Reasons for this phenomenon are discussed.

The new SAS^® High-Performance Statistics procedures were developed to respond to the growth of big data and computing capabilities. Although there is some documentation regarding how to use these new high-performance (HP) procedures, relatively little has been disseminated regarding under what specific conditions users can expect performance improvements. This paper serves as a practical guide to getting started with HP procedures in SAS^®. The paper describes the differences between key HP procedures (HPGENSELECT, HPLMIXED, HPLOGISTIC, HPNLMOD, HPREG, HPCORR, HPIMPUTE, and HPSUMMARY) and their legacy counterparts both in terms of capability and performance, with a particular focus on discrepancies in real time required to execute. Simulations were conducted to generate data sets that varied on the number of observations (10,000, 50,000, 100,000, 500,000, 1,000,000, and 10,000,000) and the number of variables (50, 100, 500, and 1,000) to create these comparisons.

Read the paper (PDF) | View the e-poster or slides (PDF)

Many organizations need to report forecasts of large numbers of time series at various levels of aggregation. Numerous model-based forecasts that are statistically generated at the lowest level of aggregation need to be combined to form an aggregate forecast that is not required to follow a fixed hierarchy. The forecasts need to be dynamically aggregated according to any subset of the time series, such as from a query. This paper proposes a technique for large-scale automatic forecast aggregation and uses SAS^® Forecast Server and SAS/ETS^® software to demonstrate this technique.

Ensemble models are a popular class of methods for combining the posterior probabilities of two or more predictive models in order to create a potentially more accurate model. This paper summarizes the theoretical background of recent ensemble techniques and presents examples of real-world applications. Examples of these novel ensemble techniques include weighted combinations (such as stacking or blending) of predicted probabilities in addition to averaging or voting approaches that combine the posterior probabilities by adding one model at a time. Fit statistics across several data sets are compared to highlight the advantages and disadvantages of each method, and process flow diagrams that can be used as ensemble templates for SAS^® Enterprise Miner™ are presented.

Motivated by the frequent need for equivalence tests in clinical trials, this presentation provides insights into tests for equivalence. We summarize and compare equivalence tests for different study designs, including one-parameter problems, designs with paired observations, and designs with multiple treatment arms. Power and sample size estimations are discussed. We also provide examples to implement the methods by using the TTEST, ANOVA, GLM, and MIXED procedures.

The experimental item response theory procedure (PROC IRT), included in recently released SAS/STAT^® 13.1 and 13.2, enables item response modeling and trait estimation in SAS^®. PROC IRT enables you to perform item parameter calibration and latent trait estimation using a wide spectrum of educational and psychological research. This paper evaluates the performance of PROC IRT in item parameter recovery and under various testing conditions. The pros and cons of PROC IRT versus BILOG-MG 3.0 are presented. For practitioners of IRT models, the development of IRT-related analysis in SAS is inspiring. This analysis offers a great choice to the growing population of IRT users. A shift to SAS can be beneficial based on several features of SAS: its flexibility in data management, its power in data analysis, its convenient output delivery, and its increasing richness in graphical presentation. It is critical to ensure the quality of item parameter calibration and trait estimation before you can continue with other components, such as test scoring, test form constructions, IRT equatings, and so on.

Scotiabank Colombian division - Colpatria, is the national leader in terms of providing credit cards, with more than 1,600,000 active cards--the equivalent to a portfolio of 700 million dollars approximately. The behavior score is used to offer credit cards through a cross-sell process, which happens only if customers have completed six months on books after using their first product with the bank. This is the minimum period of time requested by the behavior Artificial Neural Network (ANN) model. The six months on books internal policy suggests that the maturation of the client in this period is adequate, but this has never been proven. The following research aims to evaluate this hypothesis and calculate the appropriate time to offer cross-sales to new customers using Logistic Regression (Logit), while also segmenting these sales targets by their level of seniority using Discrete-Time Markov Chains (DTMC).

SAS^® In-Memory Statistics uses a powerful interactive programming interface for analytics, aimed squarely at the data scientist. We show how the custom tasks that you can create in SAS^® Studio (a web-based programming interface) can make everyone a data scientist! We explain the Common Task Model of SAS Studio, and we build a simple task in steps that carries out the basic functionality of the IMSTAT procedure. This task can then be shared amongst all users, empowering everyone on their journey to becoming a data scientist. During the presentation, it will become clear that not only can shareable tasks be created but the developer does not have to understand coding in Java, JavaScript, or ActionScript. We also use the task we created in the Visual Programming perspective in SAS Studio.

Generalized linear models (GLMs) are commonly used to model rating factors in insurance pricing. The integration of territory rating and geospatial variables poses a unique challenge to the traditional GLM approach. Generalized additive models (GAMs) offer a flexible alternative based on GLM principles with a relaxation of the linear assumption. We explore two approaches for incorporating geospatial data in a pricing model using a GAM-based framework. The ability to incorporate new geospatial data and improve traditional approaches to territory ratemaking results in further market segmentation and a better match of price to risk. Our discussion highlights the use of the high-performance GAMPL procedure, which is new in SAS/STAT^® 14.1 software. With PROC GAMPL, we can incorporate the geographic effects of geospatial variables on target loss outcomes. We illustrate two approaches. In our first approach, we begin by modeling the predictors as regressors in a GLM, and subsequently model the residuals as part of a GAM based on location coordinates. In our second approach, we model all inputs as covariates within a single GAM. Our discussion compares the two approaches and demonstrates visualization of model outputs.

The use of logistic models for independent binary data has relied first on asymptotic theory and later on exact distributions for small samples, as discussed by Troxler, Lalonde, and Wilson (2011). While the use of logistic models for dependent analysis based on exact analyses is not common, it is usually presented in the case of one-stage clustering. We present a SAS^® macro that allows the testing of hypotheses using exact methods in the case of one-stage and two-stage clustering for small samples. The accuracy of the method and the results are compared to results obtained using an R program.

Injury severity describes the severity of the injury to the person involved in the crash. Understanding the factors that influence injury severity can be helpful in designing mechanisms to reduce accident fatalities. In this research, we model and analyze the data as a hierarchy with three levels to answer the question what road, vehicle and driver-related factors influence injury severity. In this study, we used hierarchical linear modeling (HLM) for analyzing nested data from Fatality Analysis Reporting System (FARS). The results show that driver-related factors are directly related to injury severity. On the other hand, road conditions and vehicle characteristics have significant moderation impact on injury severity. We believe that our study has important policy implications for designing customized mechanisms specific to each hierarchical level to reduce the occurrence of fatal accidents.

The Armed Conflict Location & Event Data Project (ACLED) is a comprehensive public collection of conflict data for developing states. As such, the data is instrumental in informing humanitarian and development work in crisis and conflict-affected regions. The ACLED project currently manually codes for eight types of events from a summarized event description, which is a time-consuming process. In addition, when determining the root causes of the conflict across developing states, these fixed event types are limiting. How can researchers get to a deeper level of analysis? This paper showcases a repeatable combination of exploratory and classification-based text analytics provided by SAS^® Contextual Analysis, applied to the publicly available ACLED for African states. We depict the steps necessary to determine (using semi-automated processes) the next deeper level of themes associated with each of the coded event types; for example, while there is a broad protests and riots event type, how many of those are associated individually with rallies, demonstrations, marches, strikes, and gatherings? We also explore how to use predictive models to highlight the areas that require aid next. We prepare the data for use in SAS^® Visual Analytics and SAS^® Visual Statistics using SAS^® DS2 code and enable dynamic explorations of the new event sub-types. This ultimately provides the end-user analyst with additional layers of data that can be used to detect trends and deploy humanitarian aid where limited resources are needed the most.

Recent years have seen the birth of a powerful tool for companies and scientists: the Google Ngram data set, built from millions of digitized books. It can be, and has been, used to learn about past and present trends in the use of words over the years. This is an invaluable asset from a business perspective, mostly because of its potential application in marketing. The choice of words has a major impact on the success of a marketing campaign and an analysis of the Google Ngram data set can validate or even suggest the choice of certain words. It can also be used to predict the next buzzwords in order to improve marketing on social media or to help measure the success of previous campaigns. The Google Ngram data set is a gift for scientists and companies, but it has to be used with a lot of care. False conclusions can easily be drawn from straightforward analysis of the data. It contains only a limited number of variables, which makes it difficult to extract valuable information from it. Through a detailed example, this paper shows that it is essential to account for the disparity in the genre of the books used to construct the data set. This paper argues that for the years after 1950, the data set has been constructed using a much higher proportion of scientific books than for the years before. An ingenious method is developed to approximate, for each year, this unknown proportion of books coming from the scientific literature. A statistical model accounting for that change in proportion is then presented. This model is used to analyze the trend in the use of common words of the scientific literature in the 20th century. Results suggest that a naive analysis of the trends in the data can be misleading.

Road safety is a major concern for all United States of America citizens. According to the National Highway Traffic Safety Administration, 30,000 deaths are caused by automobile accidents annually. Oftentimes fatalities occur due to a number of factors such as driver carelessness, speed of operation, impairment due to alcohol or drugs, and road environment. Some studies suggest that car crashes are solely due to driver factors, while other studies suggest car crashes are due to a combination of roadway and driver factors. However, other factors not mentioned in previous studies may be contributing to automobile accident fatalities. The objective of this project was to identify the significant factors that lead to multiple fatalities in the event of a car crash.

With millions of users and peak traffic of thousands of requests a second for complex user-specific data, fantasy football offers many data design challenges. Not only is there a high volume of data transfers, but the data is also dynamic and of diverse types. We need to process data originating on the stadium playing field and user devices and make it available to a variety of different services. The system must be nimble and must produce accurate and timely responses. This talk discusses the strategies employed by and lessons learned from one of the primary architects of the National Football League's fantasy football system. We explore general data design considerations with specific examples of high availability, data integrity, system performance, and some other random buzzwords. We review some of the common pitfalls facing large-scale databases and the systems using them. And we cover some of the tips and best practices to take your data-driven applications from fantasy to reality.

Logistic regression models are commonly used in direct marketing and consumer finance applications. In this context the paper discusses two topics about the fitting and evaluation of logistic regression models. Topic 1 is a comparison of two methods for finding multiple candidate models. The first method is the familiar best subsets approach. Best subsets is then compared to a proposed new method based on combining models produced by backward and forward selection plus the predictors considered by backward and forward. This second method uses the SAS^® procedure HPLOGISTIC with selection of models by Schwarz Bayes criterion (SBC). Topic 2 is a discussion of model evaluation statistics to measure predictive accuracy and goodness-of-fit in support of the choice of a final model. Base SAS^® and SAS/STAT^® software are used.

Read the paper (PDF) | Download the data file (ZIP)

Hierarchical nonlinear mixed models are complex models that occur naturally in many fields. The NLMIXED procedure's ability to fit linear or nonlinear models with standard or general distributions enables you to fit a wide range of such models. SAS/STAT^® 13.2 enhanced PROC NLMIXED to support multiple RANDOM statements, enabling you to fit nested multilevel mixed models. This paper uses an example to illustrate the new functionality.

The popular MIXED, GLIMMIX, and NLMIXED procedures in SAS/STAT^® software fit linear, generalized linear, and nonlinear mixed models, respectively. These procedures take the classical approach of maximizing the likelihood function to estimate model. The flexible MCMC procedure in SAS/STAT can fit these same models by taking a Bayesian approach. Instead of maximizing the likelihood function, PROC MCMC draws samples (using a variety of sampling algorithms) to approximate the posterior distributions of model parameters. Similar to the mixed modeling procedures, PROC MCMC provides estimation, inference, and prediction. This paper describes how to use the MCMC procedure to fit Bayesian mixed models and compares the Bayesian approach to how the classical models would be fit with the familiar mixed modeling procedures. Several examples illustrate the approach in practice.

Analytics is now pervasive in our society, informing millions of decisions every year, and even making decisions without human intervention in automated systems. The implicit assumption is that all completed analytical projects are accurate and successful, but the reality is that some are not. Bad decisions that result from faulty analytics can be costly, embarrassing and even dangerous. By understanding the root causes of analytical failures, we can minimize the chances of repeating the mistakes of the past.

We introduce age-period-cohort (APC) models, which analyze data in which performance is measured by age of an account, account open date, and performance date. We demonstrate this flexible technique with an example from a recent study that seeks to explain the root causes of the US mortgage crisis. In addition, we show how APC models can predict website usage, retail store sales, salesperson performance, and employee attrition. We even present an example in which APC was applied to a database of tree rings to reveal climate variation in the southwestern United States.

SAS^® Forecast Server provides an excellent tool towards forecasting of time series across a variety of scenarios and industries. Given sufficient data, SAS Forecast Server can provide insights into seasonality, trend, and effects of input variables and events. In some business cases, there might not be sufficient data within individual series to produce these complex models. This paper presents an approach to forecasting these short time series. Using sufficiently long time series as a training data set, forecasted shape and volumes are created through clustering and attribute-based regressions, allowing for new series to have forecasts based on their characteristics. These forecasts are passed into SAS Forecast Server through preprocessing. Production of artificial history and input variables, matched with a custom model addition to the standard model repository, allows for future values of the artificial variable to be realized as the forecast. This process not only relieves the need for manual entry of overrides, but also allows SAS Forecast Server to choose alternative models as additional observations are provided through incremental loads.

The importance of understanding terrorism in the United States assumed heightened prominence in the wake of the coordinated attacks of September 11, 2001. Yet, surprisingly little is known about the frequency of attacks that might happen in the future and the factors that lead to a successful terrorist attack. This research is aimed at forecasting the frequency of attacks per annum in the United States and determining the factors that contribute to an attack's success. Using the data acquired from the Global Terrorism Database (GTD), we tested our hypothesis by examining the frequency of attacks per annum in the United States from 1972 to 2014 using SAS^® Enterprise Miner™ and SAS^® Forecast Studio. The data set has 2,683 observations. Our forecasting model predicts that there could be as many as 16 attacks every year for the next 4 years. From our study of factors that contribute to the success of a terrorist attack, we discovered that attack type, weapons used in the attack, place of attack, and type of target play pivotal roles in determining the success of a terrorist attack. Results reveal that the government might be successful in averting assassination attempts but might fail to prevent armed assaults and facilities or infrastructure attacks. Therefore, additional security might be required for important facilities in order to prevent further attacks from being successful. Results further reveal that it is possible to reduce the forecasted number of attacks by raising defense spending and by putting an end to the raging war in the Middle East. We are currently working on gathering data to do in-depth analysis at the monthly and quarterly level.

Analysts find the standard linear regression and analysis-of-variance models to be extremely convenient and useful tools. The standard linear model equation form is observations = (sum of explanatory variables) + residual with the assumptions of normality and homogeneity of variance. However, these tools are unsuitable for non-normal response variables in general. Using various transformations can stabilize the variance. These transformations are often ineffective because they fail to address the skewness problem. It can be complicated to transform the estimates back to their original scale and interpret the results of the analysis. At this point, we reach the limits of the standard linear model. This paper introduces generalized linear models (GzLM) using a systematic approach to adapting linear model methods on non-normal data. Why GzLM? Generalized linear models have greater power to identify model effects as statistically significant when the data are not normally distributed (Stroup, xvii). Unlike the standard linear model, the generalized linear model contains the distribution of the observations, the linear predictor or predictors, the variance function, and the link function. A few examples show how to build a GzLM for a variety of response variables that follows a Poisson, Negative Binomial, Exponential, or Gamma distribution. The SAS/STAT^® GENMOD procedure is used to compute basic analyses.

There are many methods to randomize participants in randomized control trials. If it is important to have approximately balanced groups throughout the course of the trial, simple randomization is not a suitable method. Perhaps the most common alternative method that provides balance is the blocked randomization method. A less well-known method called the treatment adaptive randomized design also achieves balance. This paper shows you how to generate an entire randomization sequence to randomize participants in a two-group clinical trial using the adaptive biased coin randomization design (ABCD), prior to recruiting any patients. Such a sequence could be used in a central randomization server. A unique feature of this method allows the user to determine the extent to which imbalance is permitted to occur throughout the sequence while retaining the probabilistic nature that is essential to randomization. Properties of sequences generated by the ABCD approach are compared to those generated by simple randomization, a variant of simple randomization that ensures balance at the end of the sequence, and by blocked randomization.

Since its inception, SAS^® Text Miner has used the singular value decomposition (SVD) to convert a term-document matrix to a representation that is crucial for building successful supervised and unsupervised models. In this presentation, using SAS^® code and SAS Text Miner, we compare these models with those that are based on SVD representations of subcomponents of documents. These more granular SVD representations are also used to provide further insights into the collection. Examples featuring visualizations, discovery of term collocations, and near-duplicate subdocument detection are shown.

Since Atul Gawande popularized the term in describing the work of Dr. Jeffrey Brenner in a New Yorker article, hot-spotting has been used in health care to describe the process of identifying super-utilizers of health care services, then defining intervention programs to coordinate and improve their care. According to Brenner's data from Camden, New Jersey, 1% of patients generate 30% of payments to hospitals, while 5% of patients generate 50% of payments. Analyzing administrative health care claims data, which contains information about diagnoses, treatments, costs, charges, and patient sociodemographic data, can be a useful way to identify super-users, as well as those who may be receiving inappropriate care. Both groups can be targeted for care management interventions. In this paper, techniques for patient outlier identification and prioritization are discussed using examples from private commercial and public health insurance claims data. The paper also describes techniques used with health care claims data to identify high-risk, high-cost patients and to generate analyses that can be used to prioritize patients for various interventions to improve their health.

You can use annotation, modify templates, and change dynamic variables to customize graphs in SAS^®. Standard graph customization methods include template modification (which most people use to modify graphs that analytical procedures produce) and SG annotation (which most people use to modify graphs that procedures such as PROC SGPLOT produce). However, you can also use SG annotation to modify graphs that analytical procedures produce. You begin by using an analytical procedure, ODS Graphics, and the ODS OUTPUT statement to capture the data that go into the graph. You use the ODS document to capture the values that the procedure sets for the dynamic variables, which control many of the details of how the graph is created. You can modify the values of the dynamic variables, and you can modify graph and style templates. Then you can use PROC SGRENDER along with the ODS output data set, the captured or modified dynamic variables, the modified templates, and SG annotation to create highly customized graphs. This paper shows you how and provides examples.

Read the paper (PDF) | Download the data file (ZIP)

Electric load forecasting is a complex problem that is linked with social activity considerations and variations in weather and climate. Furthermore, electric load is one of only a few goods that require the demand and supply to be balanced at almost real time (that is, there is almost no inventory). As a result, the utility industry could be much more sensitive to forecast error than many other industries. This electric load forecasting problem is even more challenging for holidays, which have limited historical data. Because of the limited holiday data, the forecast error for holidays is higher, on average, than it is for regular days. Identifying and using days in the history that are similar to holidays to help model the demand during holidays is not new for holiday demand forecasting in many industries. However, the electric demand in the utility industry is strongly affected by the interaction of weather conditions and social activities, making the problem even more dynamic. This paper describes an investigation into the various technologies that are used to identify days that are similar to holidays and using those days for holiday demand forecasting in the utility industry.

Contemporary data-collection processes usually involve recording information about the geographic location of each observation. This geospatial information provides modelers with opportunities to examine how the interaction of observations affects the outcome of interest. For example, it is likely that car sales from one auto dealership might depend on sales from a nearby dealership either because the two dealerships compete for the same customers or because of some form of unobserved heterogeneity common to both dealerships. Knowledge of the size and magnitude of the positive or negative spillover effect is important for creating pricing or promotional policies. This paper describes how geospatial methods are implemented in SAS/ETS^® and illustrates some ways you can incorporate spatial data into your modeling toolkit.

Managing large data sets comes with the task of providing a certain level of quality assurance, no matter what the data is used for. We present here the fundamental SAS^® procedures to perform when determining the completeness of a data set. Even though each data set is unique and has its own variables that need more examination in detail, it is important to first examine the size, time, range, interactions, and purity (STRIP) of a data set to determine its readiness for any use. This paper covers first steps you should always take, regardless of whether you're dealing with health, financial, demographic, or environmental data.

Read the paper (PDF) | Watch the recording

Looking for new ways to improve your business? Try mining your own data! Event log data is a side product of information systems generated for audit and security purposes and is seldom analyzed, especially in combination with business data. Along with the cloud computing era, more event log data has been accumulated and analysts are searching for innovative ways to take advantage of all data resources in order to get valuable insights. Process mining, a new field for discovering business patterns from event log data, has recently proved useful for business applications. Process mining shares some algorithms with data mining but it is more focused on interpretation of the detected patterns rather than prediction. Analysis of these patterns can lead to improvements in the efficiency of common existing and planned business processes. Through process mining, analysts can uncover hidden relationships between resources and activities and make changes to improve organizational structure. This paper shows you how to use SAS^® Analytics to gain insights from real event log data.

Statistical quality improvement is based on understanding process variation, which falls into two categories: variation that is natural and inherent to a process, and unusual variation due to specific causes that can be addressed. If you can distinguish between natural and unusual variation, you can take action to fix a broken process and avoid disrupting a stable process. A control chart is a tool that enables you to distinguish between the two types of variation. In many health care activities, carefully designed processes are in place to reduce variation and limit adverse events. The types of traditional control charts that are designed to monitor defect counts are not applicable to monitoring these rare events, because these charts tend to be overly sensitive, signaling unusual variation each time an event occurs. In contrast, specialized rare events charts are well suited to monitoring low-probability events. These charts have gained acceptance in health care quality improvement applications because of their ease of use and their suitability for processes that have low defect rates. The RAREEVENTS procedure, which is new in SAS/QC^® 14.1, produces rare events charts. This paper presents an overview of PROC RAREEVENTS and illustrates how you can use rare events charts to improve health care quality.

Memory-based reasoning (MBR) is an empirical classification method that works by comparing cases in hand with similar examples from the past and then applying that information to the new case. MBR modeling is based on the assumptions that the input variables are numeric, orthogonal to each other, and standardized. The latter two assumptions are taken care of by principal components transformation of raw variables and using the components instead of the raw variables as inputs to MBR. To satisfy the first assumption, the categorical variables are often dummy coded. This raises issues such as increasing dimensionality and overfitting in the training data by introducing discontinuity in the response surface relating inputs and target variables. The Weight of Evidence (WOE) method overcomes this challenge. This method measures the relative response of the target for each group level of a categorical variable. Then the levels are replaced by the pattern of response of the target variable within that category. The SAS^® Enterprise Miner™ Interactive Grouping Node is used to achieve this. With this method, the categorical variables are converted into numeric variables. This paper demonstrates the improvement in performance of an MBR model when categorical variables are WOE coded. A credit screening data set obtained from SAS^® Education that comprises 25 attributes for 3000 applicants is used for this study. Three different types of MBR models were built using the SAS Enterprise Miner MBR node to check the improvement in performance. The results show the MBR model with WOE coded categorical variables is the best, based on misclassification rate. Using this data, when WOE coding is adopted, the model misclassification rate decreases from 0.382 to 0.344 while the sensitivity of the model increases from 0.552 to 0.572.

In studies where randomization is not possible, imbalance in baseline covariates (confounding by indication) is a fundamental concern. Propensity score matching (PSM) is a popular method to minimize this potential bias, matching individuals who received treatment to those who did not, to reduce the imbalance in pre-treatment covariate distributions. PSM methods continue to advance, as computing resources expand. Optimal matching, which selects the set of matches that minimizes the average difference in propensity scores between mates, has been shown to outperform less computationally intensive methods. However, many find the implementation daunting. SAS/IML^® software allows the integration of optimal matching routines that execute in R, e.g. the R optmatch package. This presentation walks through performing optimal PSM in SAS^® through implementing R functions, assessing whether covariate trimming is necessary prior to PSM. It covers the propensity score analysis in SAS, the matching procedure, and the post-matching assessment of covariate balance using SAS/STAT^® 13.2 and SAS/IML procedures.

This paper presents the use of latent class analysis (LCA) to base the identification of a set of mutually exclusive latent classes of individuals on responses to a set of categorical, observed variables. The LCA procedure, a user-defined SAS^® procedure for conducting LCA and LCA with covariates, is demonstrated using follow-up data on substance use from Monitoring the Future panel data, a nationally representative sample of high school seniors who are followed at selected time points during adulthood. The demonstration includes guidance on data management prior to analysis, PROC LCA syntax requirements and options, and interpretation of output.

The current study looks at several ways to investigate latent variables in longitudinal surveys and their use in regression models. Three different analyses for latent variable discovery are briefly reviewed and explored. The latent analysis procedures explored in this paper are PROC LCA, PROC LTA, PROC TRAJ, and PROC CALIS. The latent variables are then included in separate regression models. The effect of the latent variables on the fit and use of the regression model compared to a similar model using observed data is briefly reviewed. The data used for this study was obtained via the National Longitudinal Study of Adolescent Health, a study distributed and collected by Add Health. Data was analyzed using SAS^® 9.4. This paper is intended for any level of SAS^® user. This paper is also written to an audience with a background in behavioral science and/or statistics.

Quality issues can easily go undetected in the field for months, even years. Why? Customers are more likely to complain through their social networks than directly through customer service channels. The Internet is the closest friend for most users and focusing purely on internal data isn't going to help. Customer complaints are often handled by marketing or customer service and aren't shared with engineering. This lack of integrated vision is due to the silo mentality in organizations and proves costly in the long run. Information systems are designed with business rules to find out of spec situations or to trigger when a certain threshold of reported cases is reached before engineering is involved. Complex business problems demand sophisticated technology to deliver advanced analytical capabilities. Organizations can react only to what they know about, but unstructured data is an invaluable asset that we must hatch on. So what if you could infuse the voice of the customer directly into engineering and detect emerging issues months earlier than your existing process? Lenovo just did that. Issues were detected in time to remove problematic parts, eventually leading to massive reductions in cost and time. Lenovo observed a significant decline in call volume to its corporate call center. It became evident that customers' perception of quality relative to the market is a voice we can't afford to ignore. Lenovo Corporate Analytics unit partnered with SAS^® to build an end-to-end process that constitutes data acquisition and preparation, analysis using linguistic text analytics models, and quality control metrics for visualization dashboards. Lenovo executives consume these reports and make informed decisions. In this paper, we talk in detail about perceptual quality and why we should care about it. We also tell you where to look for perceptual quality data and best practices for starting your own perceptual quality program through the voice of customer analytics.

As a result of globalization, the durable goods market has become increasingly competitive, with market conditions that challenge profitability for manufacturers. Moreover, high material costs and the capital-intensive nature of the industry make it essential that companies understand demand signals and utilize supply chain capacity as effectively as possible. To grow and increase profitability under these challenging market conditions, a major durable goods company has partnered with SAS to streamline analysis of pricing and profitability, optimize inventory, and improve service levels to its customers. The price of a product is determined by a number of factors, such as the strategic importance of customers, supply chain costs, market conditions, and competitive prices. Offering promotions is an important part of a marketing strategy; it impacts purchasing behaviors of business customers and end consumers. This paper describes how this company developed a system to analyze product profitability and the impact of promotion on purchasing behaviors of both their business customers and end consumers. This paper also discusses how this company uses integrated demand planning and inventory optimization to manage its complex multi-echelon supply chain. The process uses historical order data to create a statistical forecast of demand, and then optimizes inventory across the supply chain to satisfy the forecast at desired service levels.

The Limit of Detection (LoD) is defined as the lowest concentration or amount of material, target, or analyte that is consistently detectable (for polymerase chain reaction [PCR] quantitative studies, in at least 95% of the samples tested). In practice, the estimation of the LoD uses a parametric curve fit to a set of panel member (PM1, PM2, PM3, and so on) data where the responses are binary. Typically, the parametric curve fit to the percent detection levels takes on the form of a probit or logistic distribution. The SAS^® PROBIT procedure can be used to fit a variety of distributions, including both the probit and logistic. We introduce the LOD_EST SAS macro that takes advantage of the SAS PROBIT procedure's strengths and returns an information-rich graphic as well as a percent detection table with associated 95% exact (Clopper-Pearson) confidence intervals for the hit rates at each level.

Read the paper (PDF) | Download the data file (ZIP)

Retailers need critical information about the expected inventory pattern over the life of a product to make pricing, replenishment, and staffing decisions. Hotels rely on booking curves to set rates and staffing levels for future dates. This paper explores a linear state space approach to understanding these industry challenges, applying the SAS/ETS^® SSM procedure. We also use the SAS/ETS SIMILARITY procedure to provide additional insight. These advanced techniques help us quantify the relationship between the current inventory level and all previous inventory levels (in the retail case). In the hospitality example, we can evaluate how current total bookings relate to historical booking levels. Applying these procedures can produce valuable new insights about the nature of the retail inventory cycle and the hotel booking curve.

Markov chain Monte Carlo (MCMC) algorithms are an essential tool in Bayesian statistics for sampling from various probability distributions. Many users prefer to use an existing procedure to code these algorithms, while others prefer to write an algorithm from scratch. We demonstrate the various capabilities in SAS^® software to satisfy both of these approaches. In particular, we first illustrate the ease of using the MCMC procedure to define a structure. Then we step through the process of using SAS/IML^® to write an algorithm from scratch, with examples of a Gibbs sampler and a Metropolis-Hastings random walk.

Macroeconomic simulation analysis provides in-depth insights to a portfolio's performance spectrum. Conventionally, portfolio and risk managers obtain macroeconomic scenarios from third parties such as the Federal Reserve and determine portfolio performance under the provided scenarios. In this paper, we propose a technique to extend scenario analysis to an unconditional simulation capturing the distribution of possible macroeconomic climates and hence the true multivariate distribution of returns. We propose a methodology that adds to the existing scenario analysis tools and can be used to determine which types of macroeconomic climates have the most adverse outcomes for the portfolio. This provides a broader perspective on value at risk measures thereby allowing more robust investment decisions. We explain the use of SAS^® procedures like VARMAX and PROC COPULA in SAS/IML^® in this analysis.

Logic model produced propensity scores have been intensively used to assist direct marketing name selections. As a result, only customers with an absolute higher likelihood to respond are mailed offers in order to achieve cost reduction. Thus, event ROI is increased. There is a fly in the ointment, however. Compared to the model building performance time window, usually 6 months to 12 months, a marketing event time period is usually much shorter. As such, this approach lacks of the ability to deselect those who have a high propensity score but are unlikely to respond to an upcoming campaign. To consider dynamically building a complete propensity model for every upcoming camping is nearly impossible. But, incorporating time to respond has been of great interest to marketers to add another dimension for response prediction enhancement. Hence, this paper presents an inventive modeling technique combining logistic regression and the Cox Proportional Hazards Model. The objective of the fusion approach is to allow a customer's shorter next to repurchase time to compensate for his or her insignificant lower propensity score in winning selection opportunities. The method is accomplished using PROC LOGISTIC, PROC LIFETEST, PROC LIFEREF, and PROC PHREG on the fusion model that is building in a SAS^® environment. This paper also touches on how to use the results to predict repurchase response by demonstrating a case of repurchase time-shift prediction on the 12-month inactive customers of a big box store retailer. The paper also shares a results comparison between the fusion approach and logit alone. Comprehensive SAS macros are provided in the appendix.

Business problems have become more stratified and micro-segmentation is driving the need for mass-scale, automated machine learning solutions. Additionally, deployment environments include diverse ecosystems, requiring hundreds of models to be built and deployed quickly via web services to operational systems. The new SAS^® automated modeling tool allows you to build and test hundreds of models across all of the segments in your data, testing a wide variety of machine learning techniques. The tool is completely customizable, allowing you transparent access to all modeling results. This paper shows you how to identify hundreds of champion models using SAS^® Factory Miner, while generating scoring web services using SAS^® Decision Manager. Immediate benefits include efficient model deployments, which allow you to spend more time generating insights that might reveal new opportunities, expose hidden risks, and fuel smarter, well-timed decisions.

It is of paramount importance for brand managers to measure and understand consumer brand associations and the mindspace their brand captures. Brands are encoded in memory on a cognitive and emotional basis. Traditionally, brand tracking has been done by surveys and feedback, resulting in a direct communication that covers the cognitive segment and misses the emotional segment. Respondents generally behave differently under observation and in solitude. In this paper, a new brand-tracking technique is proposed that involves capturing public data from social media that focuses more on the emotional aspects. For conceptualizing and testing this approach, we downloaded nearly one million tweets for three major brands--Nike, Adidas, and Reebok--posted by users. We proposed a methodology and calculated metrics (benefits and attributes) using this data for each brand. We noticed that generally emoticons are not used in sentiment mining. To incorporate them, we created a macro that automatically cleans the tweets and replaces emoticons with an equivalent text. We then built supervised and unsupervised models on those texts. The results show that using emoticons improves the efficiency of predicting the polarity of sentiments as the misclassification rate was reduced from 0.31 to 0.24. Using this methodology, we tracked the reactions that are triggered in the minds of customers when they think about a brand and thus analyzed their mind share.

Although limited to a small fraction of health care providers, the existence and magnitude of fraud in health insurance programs requires the use of fraud prevention and detection procedures. Data mining methods are used to uncover odd billing patterns in large databases of health claims history. Efficient fraud discovery can involve the preliminary step of deploying automated outlier detection techniques in order to classify identified outliers as potential fraud before an in-depth investigation. An essential component of the outlier detection procedure is the identification of proper peer comparison groups to classify providers as within-the-norm or outliers. This study refines the concept of peer comparison group within the provider category and considers the possibility of distinct billing patterns associated with medical or surgical procedure codes identifiable by the Berenson-Eggers Type of System (BETOS). The BETOS system covers all HCPCS codes (Health Care Procedure Coding System); assigns a HCPCS code to only one BETOS code; consists of readily understood clinical categories; consists of categories that permit objective assignment& (Center for Medicare and Medicaid Services, CMS). The study focuses on the specialty General Practice and involves two steps: first, the identification of clusters of similar BETOS-based billing patterns; and second, the assessment of the effectiveness of these peer comparison groups in identifying outliers. The working data set is a sample of the summary of 2012 data of physicians active in health care government programs made publicly available by the CMS through its website. The analysis uses PROC FASTCLUS, the SAS^® cubic clustering criterion approach, to find the optimal number of clusters in the data. It also uses PROC ROBUSTREG to implement a multivariate adaptive threshold outlier detection method.

Read the paper (PDF) | Download the data file (ZIP)

In 1965, nearly half of all Americans 65 and older had no health insurance. Now, 50 years later, only 2% lack health insurance. The difference, of course, is Medicare. Medicare now covers 55 million people, about 17% of the US population, and is the single largest purchaser of personal health care. Despite this success, the rising costs of health care in general and Medicare in particular have become a growing concern. Medicare policies are important not only because they directly affect large numbers of beneficiaries, payers, and providers, but also because they affect private-sector policies as well. Analyses of Medicare policies and their consequences are complicated both by the effects of an aging population that has changing cost drivers (such as less smoking and more obesity) and by different Medicare payment models. For example, the average age of the Medicare population will initially decrease as the baby-boom generation reaches eligibility, but then increase as that generation grows older. Because younger beneficiaries have lower costs, these changes will affect cost trends and patterns that need to be interpreted within the larger context of demographic shifts. This presentation examines three Medicare payment models: fee-for-service (FFS), Medicare Advantage (MA), and Accountable Care Organizations (ACOs). FFS, originally based on payment methods used by Blue Cross and Blue Shield in the mid-1960s, pays providers for individual services (for example, physicians are paid based on the fees they charge). MA is a capitated payment model in which private plans receive a risk-adjusted rate. ACOs are groups of providers who are given financial incentives for reducing cost and maintaining quality of care for specified beneficiaries. Each model has strengths and weaknesses in specific markets. We examine each model, in addition to new data sources and more recent, innovative payment models that are likely to affect future trends.

Today's financial institutions employ tens of thousands of employees globally in the execution of manually intensive processes, from transaction processing to fraud investigation. While heuristics-based solutions have widely penetrated the industry, these solutions often work in the realm of broad generalities, lacking the nuance and experience of a human decision-maker. In today's regulatory environment, that translates to operational inefficiency and overhead as employees labor to rationalize human decisions that run counter to computed predictions. This session explores options that financial services institutions have to augment and automate day-to-day decision-making, leading to improvements in consistency, accuracy, and business efficiency. The session focuses on financial services case studies, including anti-money laundering, fraud, and transaction processing, to demonstrate real-world examples of how organizations can make the transition from predictions to decisions.

It is well documented in the literature the impact of missing data: data loss (Kim & Curry, 1977) and consequently loss of power (Raaijmakers, 1999), and biased estimates (Roth, Switzer, & Switzer, 1999). If left untreated, missing data is a threat to the validity of inferences made from research results. However, the application of a missing data treatment does not prevent researchers from reaching spurious conclusions. In addition to considering the research situation at hand and the type of missing data present, another important factor that researchers should also consider when selecting and implementing a missing data treatment is the structure of the data. In the context of educational research, multiple-group structured data is not uncommon. Assessment data gathered from distinct subgroups of students according to relevant variables (e.g., socio-economic status and demographics) might not be independent. Thus, when comparing the test performance of subgroups of students in the presence of missing data, it is important to preserve any underlying group effect by applying separate multiple imputation with multiple groups before any data analysis. Using attitudinal (Likert-type) data from the Civics Education Study (1999), this simulation study evaluates the performance of multiple-group imputation and total-group multiple imputation in terms of item parameter invariance within structural equation modeling using the SAS^® procedure CALIS and item response theory using the SAS procedure MCMC.

An ad publisher like eBay has multiple ad sources to serve ads from. Specifically, for eBay's text ads, there are two ad sources, viz. eBay Commerce Network (ECN) and Google. The problem and solution we have formulated considers two ad sources. However, the solution can be extended for any number of ad sources. A study was done on performance of ECN and Google ad sources with respect to the revenue per mille (RPM, or revenue per thousand impressions) ad queries they bring while serving ads. It was found that ECN performs better for some ad queries and Google performs better for others. Thus, our problem is to optimally allocate ad traffic between the two ad sources to maximize RPM.

As any airline traveler knows, connection time is a key element of the travel experience. A tight connection time can cause angst and concern, while a lengthy connection time can introduce boredom and a longer than desired travel time. The same elements apply when constructing schedules for airline pilots. Like passengers, pilot schedules are built with connections. Delta Air Lines operates a hub and spoke system that feeds both passengers and pilots from the spoke stations and connects them through the hub stations. Pilot connection times that are tight can result in operational disruptions, whereas extended pilot connection times are inefficient and unnecessarily costly. This paper demonstrates how Delta Air Lines used SAS^® PROC REG and PROC LOGISTIC to analyze historical data in order to build operationally robust and financially responsible pilot connections.

In cooperation with the Joint Research Centre - European Commission (JRC), we have developed a number of innovative techniques to detect outliers on a large scale. In this session, we show the power of SAS/IML^® Studio as an interactive tool for exploring and detecting outliers using customized algorithms that were built from scratch. The JRC uses this for detecting abnormal trade transactions on a large scale. The outliers are detected using the Forward Search, which starts from a central subset in the data and subsequently adds observations that are close to the current subset based on regression (R-student) or multivariate (Mahalanobis distance) output statistics. The implementation of this algorithm and its applications were done in SAS/IML Studio and converted to a macro for use in the IML procedure in Base SAS^®.

Read the paper (PDF) | Download the data file (ZIP)

In our book SAS for Mixed Models, we state that 'the majority of modeling problems are really design problems.' Graduate students and even relatively experienced statistical consultants can find translating a study design into a useful model to be a challenge. Generalized linear mixed models (GLMMs) exacerbate the challenge, because they accommodate complex designs, complex error structures, and non-Gaussian data. This talk covers strategies that have proven successful in design of experiment courses and in consulting sessions. In addition, GLMM methods can be extremely useful in planning experiments. This talk discusses methods to implement precision and power analysis to help choose between competing plausible designs and to assess the adequacy of proposed designs.

Read the paper (PDF) | Download the data file (ZIP)

In recent years, big data has been in the limelight as a solution for business issues. Implementation of big data mining has begun in a variety of industries. The variety of data types and the velocity of increasing data have been astonishing, represented as structured data stored in a relational database or unstructured data (for example, text data, GPS data, image data, and so on). In the pharmaceutical industry, big data means real-world data such as Electronic Health Record, genomics data, medical imaging data, social network data, and so on. Handling these types of big data often requires the special environment infrastructure for statistical computing. Our presentation covers case study 1: IMSTAT implementation as a large-scale parallel computation environment; conversion from business issue to data science issue in pharma; case study 2: data handling and machine learning for vertical and horizontal big data by using PROC IMSTAT; the importance of the analysis result integration; and caution points of big data mining.

Today, there are 28 million small businesses, which account for 54% of all sales in the United States. The challenge is that small businesses struggle every day to accurately forecast future sales. These forecasts not only drive investment decisions in the business, but also are used in setting daily par, determining labor hours, and scheduling operating hours. In general, owners use their gut instinct. Using SAS^® provides the opportunity to develop accurate and robust models that can unlock costs for small business owners in a short amount of time. This research examines over 5,000 records from the first year of daily sales data for a start-up small business, while comparing the four basic forecasting models within SAS^® Enterprise Guide^®. The objective of this model comparison is to demonstrate how quick and easy it is to forecast small business sales using SAS Enterprise Guide. What does that mean for small businesses? More profit. SAS provides cost-effective models for small businesses to better forecast sales, resulting in better business decisions.

The North Carolina Community College System office can quickly and easily enable colleges to compare their program's success to other college programs. Institutional researchers can now spend their days quickly looking at trends, abnormalities, and other colleges, compared to spending their days digging for data to load into a Microsoft Excel spreadsheet. We look at performance measures and how programs are being graded using SAS^® Visual Analytics.

Financial institutions are working hard to optimize credit and pricing strategies at adjudication and for ongoing account and customer management. For cards and other personal lending products, there is intense competitive pressure, together with relentless revenue challenges, that creates a huge requirement for sophisticated credit and pricing optimization tools. Numerous credit and pricing optimization applications are available on the market to satisfy these needs. We present a relatively new approach that relies heavily on the effect modeling (uplift or net lift) technique for continuous target metrics--revenue, cost, losses, and profit. Examples of effect modeling to optimize the impact of marketing campaigns are known. We discuss five essential steps on the credit and pricing optimization path: (1) setting up critical credit and pricing champion/challenger tests, (2) performance measurement of specific test campaigns, (3) effect modeling, (4) defining the best effect model, and (5) moving from the effect model to the optimal solution. These steps require specific applications that are not easily available in SAS^®. Therefore, necessary tools have been developed in SAS/STAT^® software. We go through numerous examples to illustrate credit and pricing optimization solutions.

Optimization models require continuous constraints to converge. However, some real-life problems are better described by models that incorporate discontinuous constraints. A common type of such discontinuous constraints becomes apparent when a regulation-mandated diversification requirement is implemented in an investment portfolio model. Generally stated, the requirement postulates that the aggregate of investments with individual weights exceeding certain threshold in the portfolio should not exceed some predefined total within the portfolio. This format of the diversification requirement can be defined by the rules of any specific portfolio construction methodology and is commonly imposed by the regulators. The paper discusses the impact of this type of discontinuous portfolio diversification constraint on the portfolio optimization model solution process, and develops a convergent approach. The latter includes a sequence of definite series of convergent non-linear optimization problems and is presented in the framework of the OPTMODEL procedure modeling environment. The approach discussed has been used in constructing investable equity indexes.

Read the paper (PDF) | Download the data file (ZIP) | View the e-poster or slides (PDF)

Fast detection of forest fires is a great concern among environmental experts and national park managers because forest fires create economic and ecological damage and endanger human lives. For effective fire control and resource preparation, it is necessary to predict fire occurrences in advance and estimate the possible losses caused by fires. For this purpose, using real-time sensor data of weather conditions and fire occurrence are highly recommended in order to support the predicting mechanism. The objective of this presentation is to use SAS^® 9.4 and SAS^® Enterprise Miner™ 14.1 to predict the probability of fires and to figure out special weather conditions resulting in incremental burned areas in Montesinho Park forest (Portugal). The data set was obtained from the Center for Machine Learning and Intelligent Systems at the University of California, Irvine and contains 517 observations and 13 variables from January 2000 to December 2003. Support Vector Machine analyses with variable selection were performed on this data set for fire occurrence prediction with a validation accuracy of approximately 60%. The study also incorporates Incremental Response technique and hypothesis testing to estimate the increased probability of fire, as well as the extra burned area under various conditions. For example, when there is no rain, a 27% higher chance of fires and 4.8 hectares of extra burned area are recorded, compared to when there is rain.

Due to advances in medical care and the rise in living standards, life expectancy increased on average to 79 years in the US. This resulted in an increase in aging populations and increased demand for development of technologies that aid elderly people to live independently and safely. It is possible to achieve this challenge through ambient-assisted living (AAL) technologies that help elderly people live more independently. Much research has been done on human activity recognition (HAR) in the last decade. This research work can be used in the development of assistive technologies, and HAR is expected to be the future technology for e-health systems. In this research, I discuss the need to predict human activity accurately by building various models in SAS^® Enterprise Miner™ 14.1 on a free public data set that contains 165,633 observations and 19 attributes. Variables used in this research represent the metrics of accelerometers mounted on the waist, left thigh, right arm, and right ankle of four individuals performing five different activities, recorded over a period of eight hours. The target variable predicts human activity such as sitting, sitting down, standing, standing up, and walking. Upon comparing different models, the winner is an auto neural model whose input is taken from stepwise logistic regression, which is used for variable selection. The model has an accuracy of 98.73% and sensitivity of 98.42%.

Read the paper (PDF) | View the e-poster or slides (PDF)

There is a growing recognition that mental health is a vital public health and development issue worldwide. Considerable studies have reported that there are close interactions between poverty and mental illness. The association between mental illness and poverty is cyclic and negative. Impoverished people are generally more prone to mental illness and are less capable to afford treatment. Likewise, people with mental health problems are more likely to be in poverty. The availability of free mental health treatment to people in poverty is critical to break this vicious cycle. Based on this hypothesis, a model was developed based on the responses provided by mental health facilities to a federally supported survey. We examined if we can predict whether the mental health facilities will offer free treatment to the patients who cannot afford treatment costs in the United States. About a third of the 9,076 mental health facilities who responded to the survey stated that they offer free treatment to the patients incapable of paying. Different machine learning algorithms and regression models were assessed to predict correctly which facility would offer treatment at no cost. Using a neural network model in conjunction with a decision tree for input variable selection, we found that the best performing model can predict the mental health facilities with an overall accuracy of 71.49%. Sensitivity and specificity of the selected neural network model was 82.96 and 56.57% respectively. The top five most important covariates that explained the model's predictive power are: ownership of mental health facilities, whether facilities provide a sliding fee scale, the type of facility, whether facilities provide mental health treatment service in Spanish, and whether facilities offer smoking cession services. More detailed spatial and descriptive analysis of those key variables will be conducted.

Years ago, doctors advised women with autoimmune diseases such as systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) not to become pregnant for fear of maternal health. Now, it is known that healthy pregnancy is possible for women with lupus but at the expense of a higher pregnancy complication rate. The main objective of this research is to identify key factors contributing to these diseases and to predict the occurrence rate of SLE and RA in pregnant women. Based on the approach used in this study, the prediction of adverse pregnancy outcomes for women with SLE, RA, and other diseases such as diabetes mellitus (DM) and antiphospholipid antibody syndrome (APS) can be carried out. These results will help pregnant women undergo a healthy pregnancy by receiving proper medication at an earlier stage. The data set was obtained from Cerner Health Facts data warehouse. The raw data set contains 883,473 records and 85 variables such as diagnosis code, age, race, procedure code, admission date, discharge date, total charges, and so on. Analyses were carried out with two different data sets--one for SLE patients and the other for RA patients. The final data sets had 398,742 and 397,898 records each for modeling SLE and RA patients, respectively. To provide an honest assessment of the models, the data was split into training and validation using the data partition node. Variable selection techniques such as LASSO, LARS, stepwise regression, and forward regression were used. Using a decision tree, prominent factors that determine the SLE and RA occurrence rate were identified separately. Of all the predictive models run using SAS^® Enterprise Miner™ 12.3, the model comparison node identified the decision tree (Gini) as the best model with the least misclassification rate of 0.308 to predict the SLE patients and 0.288 to predict the RA patients.

In recent years, many companies are trying to understand the rare events that are very critical in the current business environment. But a data set with rare events is always imbalanced and the models developed using this data set cannot predict the rare events precisely. Therefore, to overcome this issue, a data set needs to be sampled using specialized sampling techniques like over-sampling, under-sampling, or the synthetic minority over-sampling technique (SMOTE). The over-sampling technique deals with randomly duplicating minority class observations, but this technique might bias the results. The under-sampling technique deals with randomly deleting majority class observations, but this technique might lose information. SMOTE sampling deals with creating new synthetic minority observations instead of duplicating minority class observations or deleting the majority class observations. Therefore, this technique can overcome the problems, like biased results and lost information, found in other sampling techniques. In our research, we used an imbalanced data set containing results from a thyroid test with 3,163 observations, out of which only 4.7 percent of the observations had positive test results. Using SAS^® procedures like PROC SURVERYSELECT and PROC MODECLUS, we created over-sampled, under-sampled, and the SMOTE sampled data set in SAS^® Enterprise Guide^®. Then we built decision tree, gradient boosting, and rule induction models using four different data sets (non-sampled, majority under-sampled, minority over-sampled with majority under-sampled, and minority SMOTE sampled with majority under-sampled) in SAS^® Enterprise Miner™. Finally, based on the receiver operating characteristic (ROC) index, Kolmogorov-Smirnov statistics, and the misclassification rate, we found that the models built using minority SMOTE sampled with the majority under-sampled data yields better output for this data set.

Many inquisitive minds are filled with excitement and anticipation of response every time one posts a question on a forum. This paper explores the factors that impact the response time of the first response for questions posted in the SAS^® Community forum. The factors are contributors' availability, nature of topic, and number of contributors knowledgeable for that particular topic. The results from this project help SAS^® users receive an estimated response time, and the SAS Community forum can use this information to answer several business questions such as following: What time of the year is likely to have an overflow of questions? Do specific topics receive delayed responses? Which days of the week are the community most active? To answer such questions, we built a web crawler using Python and Selenium to fetch data from the SAS Community forum, one of the largest analytics groups. We scraped over 13,443 queries and solutions starting from January 2014 to present. We also captured several query-related attributes such as the number of replies, likes, views, bookmarks, and the number of people conversing on the query. Using different tools, we analyzed this data set after clustering the queries into 22 subtopics and found interesting patterns that can help the SAS Community forum in several ways, as presented in this paper.

In early 2006, the United States experienced a housing bubble that affected over half of the American states. It was one of the leading causes of the 2007-2008 financial recession. Primarily, the overvaluation of housing units resulted in foreclosures and prolonged unemployment during and after the recession period. The main objective of this study is to predict the current market value of a housing unit with respect to fair market rent, census region, metropolitan statistical area, area median income, household income, poverty income, number of units in the building, number of bedrooms in the unit, utility costs, other costs of the unit, and so on, to determine which factors affect the market value of the housing unit. For the purpose of this study, data was collected from the Housing Affordability Data System of the US Department of Housing and Urban Development. The data set contains 20 variables and 36,675 observations. To select the best possible input variables, several variable selection techniques were used. For example, LARS (least angle regression), LASSO (least absolute shrinkage and selection operator), adaptive LASSO, variable selection, variable clustering, stepwise regression, (PCA) principal component analysis only with numeric variables, and PCA with all variables were all tested. After selecting input variables, numerous modeling techniques were applied to predict the current market value of a housing unit. An in-depth analysis of the findings revealed that the current market value of a housing unit is significantly affected by the fair market value, insurance and other costs, structure type, household income, and more. Furthermore, a higher household income and median income of an area are associated with a higher market value of a housing unit.

Read the paper (PDF) | View the e-poster or slides (PDF)

The Oklahoma State Department of Health (OSDH) conducts home visiting programs with families that need parental support. Domestic violence is one of the many screenings performed on these visits. The home visiting personnel are trained to do initial screenings; however, they do not have the extensive information required to treat or serve the participants in this arena. Understanding how demographics such as age, level of education, and household income among others, are related to domestic violence might help home visiting personnel better serve their clients by modifying their questions based on these demographics. The objective of this study is to better understand the demographic characteristics of those in the home visiting programs who are identified with domestic violence. We also developed predictive models such as logistic regression and decision trees based on understanding the influence of demographics on domestic violence. The study population consists of all the women who participated in the Children First Program of the OSDH from 2012 to 2014. The data set contains 1,750 observations collected during screening by the home visiting personnel over the two-year period. In addition, they must have completed the Demographic form as well as the Relationship Assessment form at the time of intake. Univariate and multivariate analysis has been performed to discover the influence that age, education, and household income have on domestic violence. From the initial analysis, we can see that women who are younger than 25 years old, who haven't completed high school, and who are somewhat dependent on their husbands or partners for money are most vulnerable. We have even segmented the clients based on the likelihood of domestic violence.

This session discusses challenges and considerations typically faced in credit risk scoring as well as options and practical ways to address them. No two problems are ever the same even if the general approach is clear. Every model has its own unique characteristics and creative ways to address them. Successful credit scoring modeling projects are always based on a combination of both advanced analytical techniques and data, and a deep understanding of the business and how the model will be applied. Different aspects of the process are discussed, including feature selection, reject inferencing, sample selection and validation, and model design questions and considerations.

Automation of everyday activities holds the promise of consistency, accuracy, and relevancy. When applied to business operations, the additional benefits of governance, adaptability, and risk avoidance are realized. Prescriptive analytics empowers both systems and front-line workers to take the desired company action each and every time. And with data streaming from transactional systems, from the Internet of Things (IoT), and any other source, doing the right thing with exceptional processing speed produces the responsiveness that customers depend on. This talk describes how SAS^® and Teradata are enabling prescriptive analytics in current business environments and in the emerging IoT.

Read the paper (PDF) | View the e-poster or slides (PDF)

In customer relationship management (CRM) or consumer finance it is important to predict the time of repeated events. These repeated events might be a purchase, service visit, or late payment. Specifically, the goal is to find the probability density for the time to first event, the probability density for the time to second event, and so on. Two approaches are presented and contrasted. One approach uses discrete time hazard modeling (DTHM). The second, a distinctly different approach, uses a multinomial logistic model (MLM). Both DTHM and MLM satisfy an important consistency criterion, which is to provide probability densities whose average across all customers equals the empirical population average. DTHM requires fewer models if the number of repeated events is small. MLM requires fewer models if the time horizon is small. SAS^® code is provided through a simulation in order to illustrate the two approaches.

Donorschoose.org is a nonprofit organization that allows individuals to donate directly to public school classroom projects. Teachers from public schools post a request for funding a project with a short essay describing it. Donors all around the world can look at these projects when they log in to Donorschoose.org and donate to projects of their choice. The idea is to have a personalized recommendation webpage for all the donors, which will show them the projects, which they prefer, like and love to donate. Implementing a recommender system for the DonorsChoose.org website will improve user experience and help more projects meet their funding goals. It also will help us in understanding the donors' preferences and delivering to them what they want or value. One type of recommendation system can be designed by predicting projects that will be less likely to meet funding goals, segmenting and profiling the donors and using that information for recommending right projects when the donors log in to DonorsChoose.org.

Regression is used to examine the relationship between one or more explanatory (independent) variables and an outcome (dependent) variable. Ordinary least squares regression models the effect of explanatory variables on the average value of the outcome. Sometimes, we are more interested in modeling the median value or some other quantile (for example, the 10th or 90th percentile). Or, we might wonder if a relationship between explanatory variables and outcome variable is the same for the entire range of the outcome: Perhaps the relationship at small values of the outcome is different from the relationship at large values of the outcome. This presentation illustrates quantile regression, comparing it to ordinary least squares regression. Topics covered will include: a graphical explanation of quantile regression, its assumptions and advantages, using the SAS^® QUANTREG procedure, and interpretation of the procedure's output.

Read the paper (PDF) | Watch the recording

This paper describes the merging of designed experiments and regularized regression to find the significant factors to use for a real-time predictive model. The real-time predictive model is used in a Weyerhaeuser modified fiber mill to estimate the value of several key quality characteristics. The modified fiber product is accepted or rejected based on the model prediction or the predicted values' deviation from lab tests. To develop the model, a designed experiment was needed at the mill. The experiment was planned by a team of engineers, managers, operators, lab techs, and a statistician. The data analysis used the actual values of the process variables manipulated in the designed experiment. There were instances of the set point not being achieved or maintained for the duration of the run. The lab tests of the key quality characteristics were used as the responses. The experiment was designed with JMP^® 64-bit Edition 11.2.0 and estimated the two-way interactions of the process variables. It was thought that one of the process variables was curvilinear and this effect was also made estimable. The LASSO method, as implemented in the SAS^® GLMSELECT procedure, was used to analyze the data after cleaning and validating the results. Cross validation was used as the variable selection operator. The resulting prediction model passed all the predetermined success criteria. The prediction model is currently being used real time in the mill for product quality acceptance.

The success of any marketing promotion is measured by the incremental response and revenue generated by the targeted population known as Test in comparison with the holdout sample known as Control. An unbiased random Test and Control sampling ensures that the incremental revenue is in fact driven by the marketing intervention. However, isolating the true incremental effect of any particular marketing intervention becomes increasingly challenging in the face of overlapping marketing solicitations. This paper demonstrates how a look-alike model can be applied using the GMATCH algorithm on a SAS^® platform to design a truly comparable control group to accurately measure and isolate the impact of a specific marketing intervention.

Read the paper (PDF) | Download the data file (ZIP)

As credit unions market themselves to increase their market share against the big banks, they understandably focus on gaining new members. However, they must also retain their existing members. Otherwise, the new members they gain can easily be offset by existing members who leave. Happily, by using predictive analytics as described in this paper, keeping (and further engaging) existing members can actually be much easier and less expensive than enlisting new members. This paper provides a step-by-step overview of a relatively simple but comprehensive approach to reduce member attrition. We first prepare the data for a statistical analysis. With some basic predictive analytics techniques, we can then identify those members who have the highest chance of leaving and the highest value. For each of these members, we can also identify why they would leave, thus suggesting the best way to intervene to retain them. We then make suggestions to improve the model for better accuracy. Finally, we provide suggestions to extend this approach to further engaging existing members and thus increasing their lifetime value. This approach can also be applied to many other organizations and industries. Code snippets are shown for any version of SAS^® software; they also require SAS/STAT^® software.

Abstract logistic regression is one of the widely used regression models in statistics. Its dependent variable is a discrete variable with at least two levels. It measures the relationship between categorical dependent variables and independent variables and predicts the likelihood of having the event associated with an outcome variable. Variable reduction and screening are the techniques that reduce the redundant independent variables and filter out the independent variables with less predictive power. For several reasons, variable reduction and screening are important and critical especially when there are hundreds of covariates available that could possibly fit into the model. These techniques decrease the model estimation time and they can avoid the occurrence of non-convergence and can generate an unbiased inference of the outcome. This paper reviews various approaches for variable reduction, including traditional methods starting from univariate single logistic regressions and methods based on variable clustering techniques. It also presents hands-on examples using SAS/STAT^® software and illustrates three selection nodes available in SAS^® Enterprise Miner™: variable selection node, variable clustering node, and decision tree node.

With advances in technology, the world of sports is now offering rich data sets that are of interest to statisticians. This talk concerns some research problems in various sports that are based on large data sets. In baseball, PITCHf/x data is used to help quantify the quality of pitches. From this, questions about pitcher evaluation and effectiveness are addressed. In cricket, match commentaries are parsed to yield ball-by-ball data in order to assist in building a match simulator. The simulator can then be used to investigate optimal lineups, player evaluation, and the assessment of fielding.

Read the paper (PDF) | Watch the recording

Sometimes, the relationship between an outcome (dependent) variable and the explanatory (independent) variable(s) is not linear. Restricted cubic splines are a way of testing the hypothesis that the relationship is not linear or summarizing a relationship that is too non-linear to be usefully summarized by a linear relationship. Restricted cubic splines are just a transformation of an independent variable. Thus, they can be used not only in ordinary least squares regression, but also in logistic regression, survival analysis, and so on. The range of values of the independent variable is split up, with knots defining the end of one segment and the start of the next. Separate curves are fit to each segment. Overall, the splines are defined so that the resulting fitted curve is smooth and continuous. This presentation describes when splines might be used, how the splines are defined, choice of knots, and interpretation of the regression results.

Read the paper (PDF) | Watch the recording

Value-based reimbursement is the emerging strategy in the US healthcare system. The premise of value-based care is simple in concept--high quality and low cost provides the greatest value to patients and the various parties that fund their coverage. The basic equation for value is equally simple to compute: value=quality/cost. However, there are significant challenges to measuring it accurately. Error or bias in measuring value could result in the failure of this strategy to ultimately improve the healthcare system. This session discusses various methods and issues with risk adjustment in a value-based reimbursement model. Risk adjustment is an essential tool for ensuring that fair comparisons are made when deciding what health services and health providers have high value. The goal this presentation is to give analysts an overview of risk adjustment and to provide guidance for when, why, and how to use risk adjustment when quantifying performance of health services and healthcare providers on both cost and quality. Statistical modeling approaches are reviewed and practical issues with developing and implementing the models are discussed. Real-world examples are also provided.

Create a model with SAS Contextual Analysis, deploy the model to your Hadoop cluster, run the model using map/reduce driven by SAS Code Accelerator not alongside, but in Hadoop. Abstract: Hadoop took the world by storm as a cost efficient software framework for storing data. Hadoop is more than that, it's an analytics platform. Analytics is more than just crunching numbers, it's also about gaining knowledge from your organization's unstructured data. SAS has the tools to do both. In this paper, we'll demonstrate how to access your unstructured data, automatically create a Text Analytics model, deploy that model in Hadoop and run SAS's Code Accelerator to give you insights into your big and unstructured data lying dormant in Hadoop. We'll show you how the insights gained can help you make better informed decisions for your organization.

Read the paper (PDF) | Download the data file (ZIP) | Watch the recording

This presentation describes the new SAS^® Customer Intelligence 360 solution for multi-armed bandit analysis of controlled experiments. Multi-armed bandit analysis has been in existence since the 1950s and is sometimes referred to as the K- or N-armed bandit problem. In this problem, a gambler at a row of slot machines (sometimes known as 'one-armed bandits') has to decide which machines to play, as well as the frequency and order in which to play each machine with the objective of maximizing returns. In practice, multi-armed bandits have been used to model the problem of managing research projects in a large organization. However, the same technology can be applied within the marketing space where, for example, variants of campaign creatives or variants of customer journey sub-paths are compared to better understand customer behavior by collecting their respective response rates. Distribution weights of the variants are adjusted as time passes and conversions are observed to reflect what customers are responding to, thus optimizing the performance of the campaign. The SAS Customer Intelligence 360 analytic engine collects data at regularly scheduled time intervals and processes it through a simulation engine to return an updated weighting schema. The process is automated in the digital space and led through the solution's content serving and execution engine for interactive communication channels. Both marketing gains and costs are studied during the process to illustrate the business value of multi-arm bandit analysis. SAS^® has coupled results from traditional A/B testing to feed the multi-armed bandit engine, which are presented.

Longitudinal data with time-dependent covariates is not readily analyzed as there are inherent, complex correlations due to the repeated measurements on the sampling unit and the feedback process between the covariates in one time period and the response in another. A generalized method of moments (GMM) logistic regression model (Lalonde, Wilson, and Yin 2014) is one method for analyzing such correlated binary data. While GMM can account for the correlation due to both of these factors, it is imperative to identify the appropriate estimating equations in the model. Cai and Wilson (2015) developed a SAS^® macro using SAS/IML^® software to fit GMM logistic regression models with extended classifications. In this paper, we expand the use of this macro to allow for continuous responses and as many repeated time points and predictors as possible. We demonstrate the use of the macro through two examples, one with binary response and another with continuous response.

Revolution Analytics reports more than two million R users worldwide. SAS^® has the capability to use R code, but users have discovered a slight learning curve to performing certain basic functions such as getting data from the web. R is a functional programming language while SAS is a procedural programming language. These differences create difficulties when first making the switch from programming in R to programming in SAS. However, SAS/IML^® software enables integration between the two languages by enabling users to write R code directly into SAS/IML. This paper details the process of using the SAS/IML command Submit /R and the R package XML to get data from the web into SAS/IML. The project uses public basketball data for each of the 30 NBA teams over the past 35 years, taken directly from Basketball-Reference.com. The data was retrieved from 66 individual web pages, cleaned using R functions, and compiled into a final data set composed of 48 variables and 895 records. The seamless compatibility between SAS and R provide an opportunity to use R code in SAS for robust modeling. The resulting analysis provides a clear and concise approach for those interested in pursuing sports analytics.

Coronal mass ejections (CMEs) are massive explosions of magnetic field and plasma from the Sun. While responsible for the northern lights, these eruptions can cause geomagnetic storms and cataclysmic damage to Earth's telecommunications systems and power grid infrastructures. Hence, it is imperative to construct highly accurate predictive processes to determine whether an incoming CME will produce devastating effects on Earth. One such process, called stacked generalization, trains a variety of models, or base-learners, on a data set. Then, using the predictions from the base-learners, another model is trained to learn from the metadata. The goal of this meta-learner is to deduce information about the biases from the base-learners to make more accurate predictions. Studies have shown success in using linear methods, especially within regularization frameworks, at the meta-level to combine the base-level predictions. Here, SAS^® Enterprise Miner™ 13.1 is used to reinforce the advantages of regularization via the Least Absolute Shrinkage and Selection Operator (LASSO) on this type of metadata. This work compares the LASSO model selection method to other regression approaches when predicting the occurrence of strong geomagnetic storms caused by CMEs.

This paper examines the various sampling options that are available in SAS^® through PROC SURVEYSELECT. We do not cover all of the possible sampling methods or options that PROC SURVEYSELECT features. Instead, we look at Simple Random Sampling, Stratified Random Sampling, Cluster Sampling, Systematic Sampling, and Sequential Random Sampling.

Read the paper (PDF) | View the e-poster or slides (PDF)

With the ever growing global tourism sector, the hotel industry is also flourishing by leaps and bounds, and the standards of quality services by the tourists are also increasing. In today's cyber age where many on the Internet become part of the online community, word of mouth has steadily increased over time. According to a recent survey, approximately 46% of travelers look for online reviews before traveling. A one-star rating by users influences the peer consumer more than the five-star brand that the respective hotel markets itself to be. In this paper, we do the customer segmentation and create clusters based on their reviews. The next process relates to the creation of an online survey targeting the interests of the customers of those different clusters, which helps the hoteliers identify the interests and the hospitality expectations from a hotel. For this paper, we use a data set of 4096 different hotel reviews, with a total of 60,239 user reviews.

Sampling, whether with or without replacement, is an important component of the hypothesis testing process. In this paper, we demonstrate the mechanics and outcome of sampling without replacement, where sample values are not independent. In other words, what we get in the first sample affects what we can get for the second sample, and so on. We use the popular variant of poker known as No Limit Texas Hold'em to illustrate.

Read the paper (PDF) | View the e-poster or slides (PDF)

Similarity analysis is used to classify an entire time series by type. In this method, a distance measure is calculated to reflect the difference in form between two ordered sequences, such as are found in time series data. The mutual differences between many sequences or time series can be compiled to a similarity matrix, similar in form to a correlation matrix. When cluster methods are applied to a similarity matrix, time series with similar patterns are grouped together and placed into clusters distinct from others that contain times series with different patterns. In this example, similarity analysis is used to classify supernovae by the way the amount of light they produce changes over time.

Missing data is something that we cannot always prevent. Data can be missing due to subjects' refusing to answer a sensitive question or in fear of embarrassment. Researchers often assume that their data are missing completely at random or missing at random.' Unfortunately, we cannot test whether the mechanism condition is satisfied because missing values cannot be calculated. Alternatively, we can run simulation in SAS^® to observe the behaviors of missing data under different assumptions: missing completely at random, missing at random, and being ignorable. We compare the effects from imputation methods if we assign a set of variable of interests to missing. The idea of imputation is to see how efficient substituted values in a data set affect further studies. This lets the audience decide which method(s) would be best to approach a data set when it comes to missing data.

In forecasting, there are often situations where several time series are interrelated: components of one time series can transition into and from other time series. A Markov chain forecast model may readily capture such intricacies through the estimation of a transition probability matrix, which enables a forecaster to forecast all the interrelated time series simultaneously. A Markov chain forecast model is flexible in accommodating various forecast assumptions and structures. Implementation of a Markov chain forecast model is straightforward using SAS/IML^® software. This paper demonstrates a real-world application in forecasting a community supervision caseload in Washington State. A Markov model was used to forecast five interrelated time series in the midst of turbulent caseload changes. This paper discusses the considerations and techniques in building a Markov chain forecast model at each step. Sample code using SAS/IML is provided. Anyone interested in adding another tool to their forecasting technique toolbox will find that the Markov approach is useful and has some unique advantages in certain settings.

SAS/ETS^® 14.1 delivers a substantial number of new features to researchers who want to examine causality with observational data in addition to forecasting the future. This release adds count data models with spatial effects, new linear and nonlinear models for panel data, the X13 procedure for seasonal adjustment, and many more new features. This paper highlights the many enhancements to SAS/ETS software and demonstrates how these features can help your organization increase revenue and enhance productivity.

Disease prevalence is one of the most basic measures of the burden of disease in the field of epidemiology. As an estimate of the total number of cases of disease in a given population, prevalence is a standard in public health analysis. The prevalence of diseases in a given area is also frequently at the core of governmental policy decisions, charitable organization funding initiatives, and countless other aspects of everyday life. However, all too often, prevalence estimates are restricted to descriptive estimates of population characteristics when they could have a much wider application through the use of inferential statistics. As an estimate based on a sample from a population, disease prevalence can vary based on random fluctuations in that sample rather than true differences in the population characteristic. Statistical inference uses a known distribution of this sampling variation to perform hypothesis tests, calculate confidence intervals, and perform other advanced statistical methods. However, there is no agreed-upon sampling distribution of the prevalence estimate. In cases where the sampling distribution of an estimate is unknown, statisticians frequently rely on the bootstrap re-sampling procedure first given by Efron in 1979. This procedure relies on the computational power of software to generate repeated pseudo-samples similar in structure to an original, real data set. These multiple samples allow for the construction of confidence intervals and statistical tests to make statistical determinations and comparisons using the estimated prevalence. In this paper, we use the bootstrapping capabilities of SAS^® 9.4 to compare statistically the difference between two given prevalence rates. We create a bootstrap analog to the two-sample t test to compare prevalence rates from two states despite the fact that the sampling distribution of these estimates is unknown using SAS^®.

The increasing size and complexity of data in research and business applications require a more versatile set of tools for building explanatory and predictive statistical models. In response to this need, SAS/STAT^® software continues to add new methods. This presentation takes you on a high-level tour of five recent enhancements: new effect selection methods for regression models with the GLMSELECT procedure, model selection for generalized linear models with the HPGENSELECT procedure, model selection for quantile regression with the HPQUANTSELECT procedure, construction of generalized additive models with the GAMPL procedure, and building classification and regression trees with the HPSPLIT procedure. For each of these approaches, the presentation reviews its key concepts, uses a basic example to illustrate its benefits, and guides you to information that will help you get started.

Sensors, devices, social conversation streams, web movement, and all things in the Internet of Things (IoT) are transmitting data at unprecedented volumes and rates. SAS^® Event Stream Processing ingests thousands and even hundreds of millions of data events per second, assessing both the content and the value. The benefit to organizations comes from doing something with those results, and eliminating the latencies associated with storing data before analysis happens. This paper bridges the gap. It describes how to use streaming data as a portable, lightweight micro-analytics service for consumption by other applications and systems.

Read the paper (PDF) | Download the data file (ZIP)

Big data, small data--but what about when you have no data? Survey data commonly include missing values due to nonresponse. Adjusting for nonresponse in the analysis stage might lead different analysts to use different, and inconsistent, adjustment methods. To provide the same complete data to all the analysts, you can impute the missing values by replacing them with reasonable nonmissing values. Hot-deck imputation, the most commonly used survey imputation method, replaces the missing values of a nonrespondent unit by the observed values of a respondent unit. In addition to performing traditional cell-based hot-deck imputation, the SURVEYIMPUTE procedure, new in SAS/STAT^® 14.1, also performs more modern fully efficient fractional imputation (FEFI). FEFI is a variation of hot-deck imputation in which all potential donors in a cell contribute their values. Filling in missing values is only a part of PROC SURVEYIMPUTE. The real trick is to perform analyses of the filled-in data that appropriately account for the imputation. PROC SURVEYIMPUTE also creates a set of replicate weights that are adjusted for FEFI. Thus, if you use the imputed data from PROC SURVEYIMPUTE along with the replicate methods in any of the survey analysis procedures--SURVEYMEANS, SURVEYFREQ, SURVEYREG, SURVEYLOGISTIC, or SURVEYPHREG--you can be confident that inferences account not only for the survey design, but also for the imputation. This paper discusses different approaches for handling nonresponse in surveys, introduces PROC SURVEYIMPUTE, and demonstrates its use with real-world applications. It also discusses connections with other ways of handling missing values in SAS/STAT.

Cancer is the second-leading cause of deaths in the United States. About 10% to 15% of all lung cancers are non-small cell lung cancer, but they constitute about 26.8% of cancer deaths. An efficient cancer treatment plan is therefore of prime importance for increasing the survival chances of a patient. Cancer treatments are generally given to a patient in multiple sittings and often doctors tend to make decisions based on the improvement over time. Calculating the survival chances of the patient with respect to time and determining the various factors that influence the survival time would help doctors make more informed decisions about the further course of treatment and also help patients develop a proactive approach in making choices for the treatment. The objective of this paper is to analyze the survival time of patients suffering from non-small cell lung cancer, identify the time interval crucial for their survival, and identify various factors such as age, gender, and treatment type. The performances of the models built are analyzed to understand the significance of cubic splines used in these models. The data set is from the Cerner database with 548 records and 12 variables from 2009 to 2013. The patient records with loss to follow-up are censored. The survival analysis is performed using parametric and nonparametric methods in SAS^® Enterprise Miner™ 13.1. The analysis revealed that the survival probability of a patient is high within two weeks of hospitalization and the probability of survival goes down by 60% between weeks 5 and 9 of admission. Age, gender, and treatment type play prominent roles in influencing the survival time and risk. The probability of survival for female patients decreases by 70% and 80% during weeks 6 and 7, respectively, and for male patients the probability of survival decreases by 70% and 50% during weeks 8 and 13, respectively.

A survey is often the best way to obtain information and feedback to help in program improvement. At a bare minimum, survey research comprises economics, statistics, and psychology in order to develop a theory of how surveys measure and predict important aspects of the human condition. Through healthcare-related research papers written by experts in the industry, I hypothesize a list of several factors that are important and necessary for patients during hospital visits. Through text mining using SAS^® Enterprise Miner™ 12.1, I measured tangible aspects of hospital quality and patient care by using online survey responses and comments found on the website for the National Health Service in England. I use these patient comments to determine the majority opinion and whether it correlates with expert research on hospital quality. The implications of this research are vital in comprehending our health-care system and the factors that are important in satisfying patient needs. Analyzing survey responses can help us to understand and mitigate disparities in health-care services provided to population subgroups in the United States. Starting with online survey responses can help us to understand the overall methodology and motivation needed to analyze the surveys that have been developed for departments like the Centers for Medicare and Medicaid Services (CMS).

Read the paper (PDF) | View the e-poster or slides (PDF)

In this session, we examine the use of text analytics as a tool for strategic analysis of an organization, a leader, a brand, or a process. The software solutions SAS^® Enterprise Miner™ and Base SAS^® are used to extract topics from text and create visualizations that identify the relative placement of the topics next to various business entities. We review a number of case studies that identify and visualize brand-topic relationships in the context of branding, leadership, and service quality.

Read the paper (PDF) | Watch the recording

The recent controversy regarding former Secretary Hillary Clinton's use of a non-government, privately maintained email server provides a great opportunity to analyze real-world data using a variety of analytic techniques. This email corpus is interesting because of the challenges in acquiring and preparing the data for analysis as well as the variety of analyses that can be performed, including techniques for searching, entity extraction and resolution, natural language processing for topic generation, and social network analysis. Given the potential for politically charged discussion, rest assured there will be no discussion of politics--just fact-based analysis.

All public schools in the United States require health and safety education for their students. Furthermore, almost all states require driver education before minors can obtain a driver's license. Through extensive analysis of the Fatality Analysis Reporting System data, we have concluded that from 2011-2013 an average of 12.1% of all individuals killed in a motor vehicle accident in the United States, District of Columbia, and Puerto Rico were minors (18 years or younger). Our goal is to offer insight within our analysis in order to better road safety education to prevent future premature deaths involving motor vehicles.

It's impossible to know all of SAS^® or all of statistics. There will always be some technique that you don't know. However, there are a few techniques that anyone in biostatistics should know. If you can calculate those with SAS, life is all the better. In this session you will learn how to compute and interpret a baker's dozen of these techniques, including several statistics that are frequently confused. The following statistics are covered: prevalence, incidence, sensitivity, specificity, attributable fraction, population attributable fraction, risk difference, relative risk, odds ratio, Fisher's exact test, number needed to treat, and McNemar's test. With these 13 tools in their tool chest, even nonstatisticians or statisticians who are not specialists will be able to answer many common questions in biostatistics. The fact that each of these can be computed with a few statements in SAS makes the situation all the sweeter. Bring your own doughnuts.

The HPSUMMARY procedure provides data summarization tools to compute basic descriptive statistics for variables in a SAS^® data set. It is a high-performance version of the SUMMARY procedure in Base SAS^®. Though PROC SUMMARY is popular with data analysts, PROC HPSUMMARY is still a new kid on the block. The purpose of this paper is to provide an introduction to PROC HPSUMMARY by comparing it with its well-known counterpart, PROC SUMMARY. The comparison focuses on differences in syntax, options, and performance in terms of execution time and memory usage. Sample code, outputs, and SAS log snippets are provided to illustrate the discussion. Simulated large data is used to observe the performance of the two procedures. Using SAS^® 9.4 installed on a single-user machine with four cores available, preliminary experiments that examine performance of the two procedures show that HPSUMMARY is more memory-efficient than SUMMARY when the data set is large. (For example, SUMMARY failed due to insufficient memory, whereas HPSUMMARY finished successfully). However, there is no evidence of a performance advantage of the HPSUMMARY over the SUMMARY procedures in this single-user machine.

Read the paper (PDF) | View the e-poster or slides (PDF)

This paper is a primer on the practice of designing, selecting, and making inferences on a statistical sample, where the goal is to estimate the magnitude of error in a book value total. Although the concepts and syntax are presented through the lens of an audit of health-care insurance claim payments, they generalize to other contexts. After presenting the fundamental measures of uncertainty that are associated with sample-based estimates, we outline a few methods to estimate the sample size necessary to achieve a targeted precision threshold. The benefits of stratification are also explained. Finally, we compare several viable estimators to quantify the book value discrepancy, making note of the scenarios where one might be preferred over the others.

This paper shows how SAS^® can be used to obtain a Time Series Analysis of data regarding World War II. This analysis tests whether Truman's justification for the use of atomic weapons was valid. Truman believed that by using the atomic weapons, he would prevent unacceptable levels of U.S. casualties that would be incurred in the course of a conventional invasion of the Japanese islands.

Read the paper (PDF) | View the e-poster or slides (PDF)

Inherently, mixed modeling with SAS/STAT^® procedures (such as GLIMMIX, MIXED, and NLMIXED) is computationally intensive. Therefore, considerable memory and CPU time can be required. The default algorithms in these procedures might fail to converge for some data sets and models. This encore presentation of a paper from SAS Global Forum 2012 provides recommendations for circumventing memory problems and reducing execution times for your mixed-modeling analyses. This paper also shows how the new HPMIXED procedure can be beneficial for certain situations, as with large sparse mixed models. Lastly, the discussion focuses on the best way to interpret and address common notes, warnings, and error messages that can occur with the estimation of mixed models in SAS^® software.

Read the paper (PDF) | Watch the recording

Educational systems at the district, state, and national levels all report possessing amazing student-level longitudinal data systems (LDS). Are the LDS systems improving educational outcomes for students? Are they guiding development of effective instructional practices? Are the standardized exams measuring student knowledge relative to the learning expectations? Many questions exist about the effective use of the LDS system and educational data, but data architecture and analytics (including the products developed by SAS^®) are not designed to answer any of these questions. However, the ability to develop more effective educational interfaces, improve use of data to the classroom level, and improve student outcomes, might only be available through use of SAS. The purpose of this session and paper is to demonstrate an integrated use of SAS tools to guide the transformation of data to analytics that improve educational outcomes for all students.

Machine learning is gaining popularity, fueled by the rapid advancement of computing technology. Machine learning offers the ability to continuously adjust predictive models based on real-time data, and to visualize changes and take action on the new information. Hear from two PhD's from SAS and Intel about how SAS^® machine learning, working with SAS streaming analytics and Intel microprocessor technology, makes this possible.

This paper proposes a technique to implement wavelet analysis (WA) for improving a forecasting accuracy of the autoregressive integrated moving average model (ARIMA) in nonlinear time-series. With the assumption of the linear correlation, and conventional seasonality adjustment methods used in ARIMA (that is, differencing, X11, and X12), the model might fail to capture any nonlinear pattern. Rather than directly model such a signal, we decompose it to less complex components such as trend, seasonality, process variations, and noises, using WA. Then, we use them as exogenous variables in the autoregressive integrated moving average with explanatory variable model (ARIMAX). We describe a background of WA. Then, the code and a detailed explanation of WA based on multi-resolution analysis (MRA) in SAS/IML^® software are demonstrated. The idea and mathematical basis of ARIMA and ARIMAX are also given. Next, we demonstrate our technique in forecasting applications using SAS^® Forecast Studio. The demonstrated time-series are nonlinear in nature from different fields. The results suggest that WA effects are good regressors in ARIMAX, which captures nonlinear patterns well.

Administrative health databases, including hospital and physician records, are frequently used to estimate the prevalence of chronic diseases. Disease-surveillance information is used by policy makers and researchers to compare the health of populations and develop projections about disease burden. However, not all cases are captured by administrative health databases, which can result in biased estimates. Capture-recapture (CR) models, originally developed to estimate the sizes of animal populations, have been adapted for use by epidemiologists to estimate the total sizes of disease populations for such conditions as cancer, diabetes, and arthritis. Estimates of the number of cases are produced by assessing the degree of overlap among incomplete lists of disease cases captured in different sources. Two- and three-source CR models are most commonly used, often with covariates. Two important assumptions--independence of capture in each data source and homogeneity of capture probabilities, which underlie conventional CR models--are unlikely to hold in epidemiological studies. Failure to satisfy these assumptions bias the model results. Log-linear, multinomial logistic regression, and conditional logistic regression models, if used properly, can incorporate dependency among sources and covariates to model the effect of heterogeneity in capture probabilities. However, none of these models is optimal, and researchers might be unfamiliar with how to use them in practice. This paper demonstrates how to use SAS^® to implement the log-linear, multinomial logistic regression, and conditional logistic regression CR models. Methods to address the assumptions of independence between sources and homogeneity of capture probabilities for a three-source CR model are provided. The paper uses a real numeric data set about Parkinson's disease involving physician claims, hospital abstract, and prescription drug records from one Canadian province. Advantages and disadvantages of each model are discus sed.

Health care and other programs collect large amounts of information in order to administer the program. A health insurance plan, for example, probably records every physician service, hospital and emergency department visit, and prescription medication--information that is collected in order to make payments to the various providers (hospitals, physicians, pharmacists). Although the data are collected for administrative purposes, these databases can also be used to address research questions, including questions that are unethical or too expensive to answer using randomized experiments. However, when subjects are not randomly assigned to treatment groups, we worry about assignment bias--the possibility that the people in one treatment group were healthier, smarter, more compliant, etc., than those in the other group, biasing the comparison of the two treatments. Propensity score methods are one way to adjust the comparisons, enabling research using administrative data to mimic research using randomized controlled trials. In this presentation, I explain what the propensity score is, how it is used to compare two treatments, and how to implement a propensity score analysis using SAS^®. Two methods using propensity scores are presented: matching and inverse probability weighting. Examples are drawn from health services research using the administrative databases of the Ontario Health Insurance Plan, the single payer for medically necessary care for the 13 million residents of Ontario, Canada. However, propensity score methods are not limited to health care. They can be used to examine the impact of any nonrandomized program, as long as there is enough information to calculate a credible propensity score.

Someone has aptly said, Las Vegas looks the way one would imagine heaven must look at night. What if you know the secret to run a plethora of various businesses in the entertainment capital of the world? Nothing better, right? Well, we have what you want, all the necessary ingredients for you to precisely know what business to target in a particular locality of Las Vegas. Yelp, a community portal, wants to help people finding great local businesses. They cover almost everything from dentists and hair stylists through mechanics and restaurants. Yelp's users, Yelpers, write reviews and give ratings for all types of businesses. Yelp then uses this data to make recommendations to the Yelpers about which institutions best fit their individual needs. We have the yelp academic data set comprising 1.6 million reviews and 500K tips by 366K in 61K businesses across several cities. We combine current Yelp data from all the various data sets for Las Vegas to create an interactive map that provides an overview of how a business runs in a locality and how the ratings and reviews tickers a business. We answer the following questions: Where is the most appropriate neighborhood to start a new business (such as cafes, bars, and so on)? Which category of business has the greatest total count of reviews that is the most talked about (trending) business in Las Vegas? How does a business' working hours affect the customer reviews and the corresponding rating of the business? Our findings present research for further understanding of perceptions of various users, while giving reviews and ratings for the growth of a business by encompassing a variety of topics in data mining and data visualization.

Many SAS^® procedures can be used to analyze large amounts of correlated data. This study was a secondary analysis of data obtained from the South Carolina Revenue and Fiscal Affairs Office (RFA). The data includes medical claims from all health care systems in South Carolina (SC). This study used the SAS procedure GENMOD to analyze a large amount of correlated data about Military Health Care (MHS) system beneficiaries who received behavioral health care in South Carolina Health Care Systems from 2005 to 2014. Behavioral health (BH) was defined by Major Diagnostic Code (MDC) 19 (mental disorders and diseases) and 20 (alcohol/drug use). MDCs are formed by dividing all possible principal diagnoses from the International Classification Diagnostic (ICD-9) codes into 25 mutually exclusive diagnostic categories. The sample included a total of 6,783 BH visits and 4,827 unique adult and child patients that included military service members, veterans, and their adult and child dependents who have MHS insurance coverage. PROC GENMOD included a multivariate GEE model with type of BH visit (mental health or substance abuse) as the dependent variable, and with gender, race group, age group, and discharge year as predictors. Hospital ID was used in the repeated statement with different correlation structures. Gender was significant with both independent correlation (p = .0001) and exchangeable structure (p = .0003). However, age group was significant using the independent correlation (p = .0160), but non-significant using the exchangeable correlation structure (p = .0584). SAS is a powerful statistical program for analyzing large amounts of correlated data with categorical outcomes.

Read the paper (PDF) | View the e-poster or slides (PDF)

The HPSPLIT procedure, a powerful new addition to SAS/STAT^®, fits classification and regression trees and uses the fitted trees for interpretation of data structures and for prediction. This presentation shows the use of PROC HPSPLIT to analyze a number of data sets from different subject areas, with a focus on prediction among subjects and across landscapes. A critical component of these analyses is the use of graphical tools to select a model, interpret the fitted trees, and evaluate the accuracy of the trees. The use of different kinds of cross validation for model selection and model assessment is also discussed.

Stephen Curry, James Harden, and LeBron James are considered to be three of the most gifted professional basketball players in the National Basketball Association (NBA). Each year the Kia Most Valuable Player (MVP) award is given to the best player in the league. Stephen Curry currently holds this title, followed by James Harden and LeBron James, the first two runners-up. The decision for MVP was made by a panel of judges comprised of 129 sportswriters and broadcasters, along with fans who were able to cast their votes through NBA.com. Did the judges make the correct decision? Is there statistical evidence that indicates that Stephen Curry is indeed deserving of this prestigious title over James Harden and LeBron James? Is there a significant difference between the two runners-up? These are some of the questions that are addressed through this project. Using data collected from NBA.com for the 2014-2015 season, a variety of parametric and nonparametric k-sample methods were used to test 20 quantitative variables. In an effort to determine which of the three players is the most deserving of the MVP title, post-hoc comparisons were also conducted on the variables that were shown to be significant. The time-dependent variables were standardized, because there was a significant difference in the number of minutes each athlete played. These variables were then tested and compared with those that had not been standardized. This led to significantly different outcomes, indicating that the results of the tests could be misleading if the time variable is not taken into consideration. Using the standardized variables, the results of the analyses indicate that there is a statistically significant difference in the overall performances of the three athletes, with Stephen Curry outplaying the other two players. However, the difference between James Harden and LeBron James is not so clear.

Read the paper (PDF) | View the e-poster or slides (PDF)

Most businesses have benefited from using advanced analytics for marketing and other decision making. But to apply analytical techniques to pharmaceutical marketing is challenging and emerging as it is critical to ensure that the analysis makes sense from the medical side. The drug for a specific disease finally consumed is directly or indirectly influenced by many factors, including the disease origins, health-care system policy, physicians' clinical decisions, and the patients' perceptions and behaviors. The key to pharmaceutical marketing is in identifying the targeted populations for specific diseases and to focus on those populations. Because the health-care environment consistently changes, the predictive models are important to predict the change of the targeted population over time based on the patient journey and epidemiology. Time series analysis is used to forecast the number of cases of infectious diseases; correspondingly, over the counter and prescribed medicines for the specific disease could be predicted. The accurate prediction provides valuable information for the strategic plan of campaigns. For different diseases, different analytical techniques are applied. By taking the medical features of the disease and epidemiology into account, the prediction of the potential and total addressable markets can reveal more insightful marketing trends. And by simulating the important factors and quantifying how they impact the patient journey within the typical health-care system, the most accurate demand for specific medicines or treatments could be discovered. Through monitoring the parameters in the dynamic simulation, the smart decision can be made using what-if comparisons to optimize the marketing result.

An important strength of observational studies is the ability to estimate a key behavior's or treatment's effect on a specific health outcome. This is a crucial strength as most health outcomes research studies are unable to use experimental designs due to ethical and other constraints. Keeping this in mind, one drawback of observational studies (that experimental studies naturally control for) is that they lack the ability to randomize their participants into treatment groups. This can result in the unwanted inclusion of a selection bias. One way to adjust for a selection bias is through the use of a propensity score analysis. In this study, we provide an example of how to use these types of analyses. Our concern is whether recent substance abuse has an effect on an adolescent's identification of suicidal thoughts. In order to conduct this analysis, a selection bias was identified and adjustment was sought through three common forms of propensity scoring: stratification, matching, and regression adjustment. Each form is separately conducted, reviewed, and assessed as to its effectiveness in improving the model. Data for this study was gathered through the Youth Risk Behavior Surveillance System, an ongoing nation-wide project of the Centers for Disease Control and Prevention. This presentation is designed for any level of statistician, SAS^® programmer, or data analyst with an interest in controlling for selection bias, as well as for anyone who has an interest in the effects of substance abuse on mental illness.

The increasing popularity and affordability of wearable devices, together with their ability to provide granular physical activity data down to the minute, have enabled researchers to conduct advanced studies on the effects of physical activity on health and disease. This provides statistical programmers the challenge of processing data and translating it into analyzable measures. One such measure is the number of time-specific bouts of moderate to vigorous physical activity (MVPA) (similar to exercise), which is needed to determine whether the participant meets current physical activity guidelines (for example, 150 minutes of MVPA per week performed in bouts of at least 20 minutes). In this paper, we illustrate how we used SAS^® arrays to calculate the number of 20-minute bouts of MVPA per day. We provide working code on how we processed Fitbit Flex data from 63 healthy volunteers whose physical activities were monitored daily for a period of 12 months.

Read the paper (PDF) | Download the data file (ZIP)

In many discrete-event simulation projects, the chief goal is to investigate the performance of a system. You can use output data to better understand the operation of the real or planned system and to conduct various what-if analyses. But you can also use simulation for validation--specifically, to validate a solution found by an optimization model. In optimization modeling, you almost always need to make some simplifying assumptions about the details of the system you are modeling. These assumptions are especially important when the system includes random variation--for example, in the arrivals of individuals, their distinguishing characteristics, or the time needed to complete certain tasks. A common approach holds each random element at some nominal value (such as the mean of its observed values) and proceeds with the optimization. You can do better. In order to test an optimization model and its underlying assumptions, you can build a simulation model of the system that uses the optimal solution as an input and simulates the system's detailed behavior. The simulation model helps determine how well the optimal solution holds up when randomness and perhaps other logical complexities (which the optimization model might have ignored, summarized, or modeled only approximately) are accounted for. Simulation might confirm the optimization results or highlight areas of concern in the optimization model. This paper describes cases in which you can use simulation and optimization together in this manner and discusses the positive implications of this complementary analytic approach. For the reader, prior experience with optimization and simulation is helpful but not required.

Read the paper (PDF) | Download the data file (ZIP)

Cycling is one of the fastest growing recreational activities and sports in India. Many non-government organizations (NGOs) support this emission-free mode of transportation that helps protect the environment from air pollution hazards. Lots of cyclist groups in metropolitan areas organize numerous ride events and build social networks. Although India was a bit late for joining the Internet, the social networking sites are getting popular in every Indian state and are expected to grow after the announcement of Digital India Project. Many networking groups and cycling blogs share tons of information and resources across the globe. However, these blogs posts are difficult to categorize according to their relevance, making it difficult to access required information quickly on the blogs. This paper provides ways to categorize the content of these blog posts and classify them in meaningful categories for easy identification. The initial data set is created from scraping the online cycling blog posts (for example, Cyclists.in, velocrushindia.com, and so on) using Python. The initial data set consists of 1,446 blog posts with titles and blog content since 2008. Approximately 25% of the blog posts are manually categorized into six different categories, such as Ride stories, Events, Nutrition, Bicycle hacks, Bicycle Buy and Sell, and Others, by three independent raters to generate a training data set. The common blog-post categories are identified, and the text classification model is built to identify the blog-post classification and generate the text rules to classify the blogs based on those categories using SAS^® Text Miner. To improve the look and feel of the social networking blog, a tool developed in JavaScript automatically classifies the blog posts into six categories and provides appropriate suggestions to the blog user.

Multivariate statistical analysis plays an increasingly important role as the number of variables being measured increases in educational research. In both cognitive and noncognitive assessments, many instruments that researchers aim to study contain a large number of variables, with each measured variable assigned to a specific factor of the bigger construct. Recalling the educational theories or empirical research, the factor of each instrument usually emerges the same way. Two types of factor analysis are widely used in order to understand the latent relationships among these variables based on different scenarios. (1) Exploratory factor analysis (EFA), which is performed by using the SAS^® procedure PROC FACTOR, is an advanced statistical method used to probe deeply into the relationship among the variables and the larger construct and then develop a customized model for the specific assessment. (2) When a model is established, confirmatory factor analysis (CFA) is conducted by using the SAS procedure PROC CALIS to examine the model fit of specific data and then make adjustments for the model as needed. This paper presents the application of SAS to conduct these two types of factor analysis to fulfill various research purposes. Examples using real noncognitive assessment data are demonstrated, and the interpretation of the fit statistics is discussed.

The objective of this study is to use the GLM procedure in SAS^® to solve a complex linkage problem with multiple test forms in educational research. Typically, the ABSORB option in the GLM procedure makes this task relatively easy to implement. Note that for educational assessments, to apply one-dimensional combinations of two-parameter logistic (2PL) models (Hambleton, Swaminathan, and Rogers 1991, ch. 1) and generalized partial credit models (Muraki 1997) to a large-scale high-stakes testing program with very frequent administrations requires a practical approach to link test forms. Haberman (2009) suggested a pragmatic solution of simultaneous linking to solve the challenging linking problem. In this solution, many separately calibrated test forms are linked by the use of least-squares methods. In SAS, the GLM procedure can be used to implement this algorithm by the use of the ABSORB option for the variable that specifies administrations, as long as the data are sorted by order of administration. This paper presents the use of SAS to examine the application of this proposed methodology to a simple case of real data.

The LACE readmission risk score is a methodology used by Kaiser Permanente Northwest (KPNW) to target and customize readmission prevention strategies for patients admitted to the hospital. This presentation shares how KPNW used SAS^® in combination with Epic's Datalink to integrate the LACE score into its electronic health record (EHR) for usage in real time. The LACE score is an objective measure, composed of four components including: L) length of stay; A) acuity of admission; C) pre-existing co-morbidities; and E) Emergency department (ED) visits in the prior six months. SAS was used to perform complex calculations and combine data from multiple sources (which was not possible for the EHR alone), and then to calculate a score that was integrated back into the EHR. The technical approach includes a trigger macro to kick off the process once the database ETL completes, several explicit and implicit proc SQL statements, a volatile temp table for filtering, and a series of SORT, MEANS, TRANSPOSE, and EXPORT procedures. We walk through the technical approach taken to generate and integrate the LACE score into Epic, as well as describe the challenges we faced, how we overcame them, and the beneficial results we have gained throughout the process.

In health care and epidemiological research, there is a growing need for basic graphical output that is clear, easy to interpret, and easy to create. SAS^® 9.3 has a very clear and customizable graphic called a Radar Graph, yet it can only display the unique responses of one variable and would not be useful for multiple binary variables. In this paper we describe a way to display multiple binary variables for a single population on a single radar graph. Then we convert our method into a macro with as few parameters as possible to make this procedure available for everyday users.

Read the paper (PDF) | Download the data file (ZIP)

Average yards per reception, as well as number of touchdowns, are commonly used to rank National Football League (NFL) players. However, scoring touchdowns lowers the player's average since it stops the play and therefore prevents the player from gaining more yardage. If yardages are tabulated in a life table, then yardage from a touchdown play is denoted as a right-censored observation. This paper discusses the application of the SAS/STAT^® Kaplan-Meier product-limit estimator to adjust these averages. Using 15 seasons of NFL receiving data, the relationship between touchdown rates and average yards per reception is compared, before and after adjustment. The modification of adjustments when a player incurred a 2-point safety during the season is also discussed.

Because optimization models often do not capture some important real-world complications, a collection of optimal or near-optimal solutions can be useful for decision makers. This paper uses various techniques for finding the k best solutions to the linear assignment problem in order to illustrate several features recently added to the OPTMODEL procedure in SAS/OR^® software. These features include the network solver, the constraint programming solver (which can produce multiple solutions), and the COFOR statement (which allows parallel execution of independent solver calls).

Read the paper (PDF) | Download the data file (ZIP)

The SURVEYSELECT procedure is useful for sample selection for a wide variety of applications. This paper presents the application of PROC SURVEYSELECT to a complex scenario involving a state-wide educational testing program, namely, drawing interdependent stratified cluster samples of schools for the field-testing of test questions, which we call items. These stand-alone field tests are given to only small portions of the testing population and as such, a stratified procedure was used to ensure representativeness of the field-test samples. As the field test statistics for these items are evaluated for use in future operational tests, an efficient procedure is needed to sample schools, while satisfying pre-defined sampling criteria and targets. This paper provides an adaptive sampling application and then generalizes the methodology, as much as possible, for potential use in more industries.

Read the paper (PDF) | View the e-poster or slides (PDF)

Each night on the news we hear the level of the Dow Jones Industrial Average along with the 'first difference,' which is today's price-weighted average minus yesterday's. It is that series of first differences that excites or depresses us each night as it reflects whether stocks made or lost money that day. Furthermore, the differences form the data series that has the most addressable statistical features. In particular, the differences have the stationarity requirement, which justifies standard distributional results such as asymptotically normal distributions of parameter estimates. Differencing arises in many practical time series because they seem to have what are called 'unit roots,' which mathematically indicate the need to take differences. In 1976, Dickey and Fuller developed the first well-known tests to decide whether differencing is needed. These tests are part of the ARIMA procedure in SAS/ETS^® in addition to many other time series analysis products. I'll review a little of what is was like to do the development and the required computing back then, say a little about why this is an important issue, and focus on examples.

Read the paper (PDF) | Watch the recording

SAS/IML^® 14.1 enables you to author, install, and call packages. A package consists of SAS/IML source code, documentation, data sets, and sample programs. Packages provide a simple way to share SAS/IML functions. An expert who writes a statistical analysis in SAS/IML can create a package and upload it to the SAS/IML File Exchange. A nonexpert can download the package, install it, and immediately start using it. Packages provide a standard and uniform mechanism for sharing programs, which benefits both experts and nonexperts. Packages are very popular with users of other statistical software, such as R. This paper describes how SAS/IML programmers can construct, upload, download, and install packages. They're not wrapped in brown paper or tied up with strings, but they'll soon be a few of your favorite things!