This paper, by SAS Education instructor extraordinaire and Senior Manager Cat Truxillo, demonstrates the new case-level residuals in the CALIS procedure and how they differ from classic residuals in structural equation modeling (SEM). Residual analysis has a long history in statistical modeling of finding unusual observations in the sample data. However, in SEM, case-level residuals are considerably more difficult to define because of (1) latent variables in the analysis and (2) the multivariate nature of these models. Historically, residual analysis in SEM has been confined to residuals obtained as the difference between the sample and model-implied covariance matrices. Enhancements to the CALIS procedure enable users to obtain case-level residuals as well. This enables a more complete residual and influence analysis. Several examples showing mean/covariance residuals and case-level residuals are presented.

You might have attended a workshop by Distinguished Research Statistician Developer Warren Kuhfeld on statistical graphics using ODS. More recent workshops have focused on advanced methods, and Kuhfeld has put the material into book form. These examples illustrate the use of the Graph Template Language, PROC SGPLOT, and SG annotation. They are a great source of information for users who understand the basics of ODS Graphics and want to improve their skills. For more information, see the blog post.

The use of Bayesian methods has become increasingly popular in modern statistical analysis, with applications in numerous scientific fields. In recent releases, SAS has provided a wealth of tools for Bayesian analysis, with convenient access through several popular procedures as well as the MCMC procedure, which is designed for general Bayesian modeling. This paper introduces the principles of Bayesian inference and reviews the steps in a Bayesian analysis. It then describes the built-in Bayesian capabilities provided in SAS/STAT, which became available for all platforms in SAS/STAT 9.3, with examples from the GENMOD and PHREG procedures. How to specify prior distributions, evaluate convergence diagnostics, and interpret the posterior summary statistics is also discussed.

We thought we'd highlight some of our own division's talks at SAS Global Forum, giving you the inside scoop on what we think you should attend if you are going, and what proceedings papers you should look out for if you're not. In addition, don't forget the Super Demos, short presentations in the software demonstration area that provide important information about key software features. This year we have 28 Super Demos in the Analytics area alone, and 126 overall. We'll convert many of the analytical demos into web videos during the summer. Added this year are some paper suggestions from the R&D staff. Keep looking, as information is being added all the time.

Not too many statisticians, especially biostatisticians, go far before encountering longitudinal data, and depending on your context, you have a choice of analysis methods. These videos teach you about the structural equation modeling (SEM) approach to longitudinal data analysis using the CALIS procedure, the method of functional modeling of longitudinal data using the SSM procedure in SAS/ETS® software, and coping with dropouts in longitudinal data with the GEE procedure.

Distinguished Research Statistician Warren Kuhfeld explains how to create a fit plot with PROC REG and how to use a categorical variable to group points. Examples show how to change the data object that underlies a graph and how to change how parts of the graph are labeled or displayed.

Many of you are upgrading to SAS 9.4 and the 14.1 analytical products releases. Sure, you can review the "What's New" chapter (and you should) or read the paper "SAS/STAT 14.1: Methods for Massive, Missing, or Multifaceted Data”, but you could also watch the SAS/STAT developers tell you what they think are the most important new features in SAS/STAT software.

Traditional Shewhart charts that are designed to monitor defect counts are not applicable to monitoring rare events. Such charts tend to be overly sensitive, signaling unusual variation each time an event occurs. This problem can be mitigated by increasing the subgroup sample size, but at the cost of delaying data analysis. In contrast, a rare events chart is well suited to monitoring low-probability events. Learn about the RAREEVENTS procedure, new in SAS/QC® 14.1, which produces rare events charts. These charts have now gained acceptance in health-care quality improvement applications.

Distinguished Research Statistician Rick Wicklin discusses four essential sampling methods in SAS that are available in the SURVEYSELECT procedure. Many simulation and resampling tasks use one of these methods. In addition, Wicklin describes four ways of creating design matrices in SAS. Usually, you let the modeling procedure worry about the design matrix, which you direct with your choice of terms in MODEL and CLASS statements. But sometimes you need to create your own design matrix for various reasons, such as for use in a procedure such as PROC MCMC or PROC NLMIXED or in your own SAS/IML® program.

Warren Kuhfeld will speak on annotating graphs from analytical procedures at the PharmaSUG Conference, May 8–11 in Denver.  See more.

Modeling Continuous Proportions: Normal and Beta Regression Models

Modeling Continuous Proportions: Fractional and 4- (or 5-) Parameter Logit Models

When you model response data consisting of proportions (or percentages), the observed values can be continuous or represent a summarized (or aggregated) binary response. Models for such data assume that the proportions represent a set of independent Bernoulli trials and have a binomial distribution. However, if the values are proportions of area covered or affected by some agent, or are proportions of a mixture, then they do not represent a set of trials and might not have a binomial distribution. Examples include the concentration of a drug in the blood, the proportion of area damaged in a forest fire, and the proportion of a chemical that is absorbed. Modeling approaches for such data include assuming the proportions are from normal or beta distributions. An alternative approach that does not require specification of a distribution for the proportions is the fractional logistic model that uses a quasi-likelihood function for estimation. These modeling approaches for continuous proportions are illustrated using the GLIMMIX, NLMIXED, and NLIN procedures.