- Capabilities
- Enhancements
Ten Great Reasons to Upgrade to SAS/ETS® 9.2
Now that the full release of SAS 9.2 is available, many customer sites are making decisions about when to install this new release. Here are the top ten reasons why we think that econometricians should update to SAS 9.2 right away!
ODS Statistical Graphics
ODS Statistical Graphics changes everything about how you create statistical graphs in SAS software. Many SAS/ETS procedures now take advantage of ODS graphics, producing appropriate graphs as part of the statistical output. Plots are produced automatically when the ODS Graphics feature is on, and additional plots can be requested with the PLOTS= option. You can make ad hoc changes with the ODS Graphics Editor, or you can permanently customize graphs by modifying the underlying graph templates. In addition, SAS/GRAPH® 9.2 provides a number of new procedures designed specifically for data analysis.
Previously experimental procedures now production status
The PANEL procedure, for analysis of panel data structures, and the COUNTREG procedure, for regression where the dependent variable is an event count, have been improved and are now production status.
Time series similarity analysis
Time series similarity analysis measures the similarity between pairs of time series, and can be used for time series clustering, classification, or search. In SAS 9.2, the new SIMILARITY procedure computes similarity measures between input sequences and target sequences, and can “slide” the target sequence with respect to the input to account for varying lags in the occurrence of similar patterns. An important use of PROC SIMILARITY is clustering of time series into groups that have similar patterns for time series data mining and for new-product forecasting applications.
Automated forecasting that uses exponential smoothing methods
The new ESM procedure provides forecasting by using automated exponential smoothing. It provides a quick way to generate forecasts for many time series or for transactional data in one step. The ESM procedure serves as a successor to the older FORECAST procedure.
Multivariate GARCH models
The generalized autoregressive conditional heteroscedasticity (GARCH) model does for the time-varying volatility of a variable what ARIMA models do for the conditional mean of the variable. By using the new GARCH statement in the VARMAX procedure, you can combine a multivariate model for the time-varying covariance matrix of the vector of dependent time series with vector autoregressive and vector transfer function models (VARMAX model) for the time-varying mean.
Stochastic frontier models
To be able to evaluate a firm's technological efficiency, you need to know a production function of a fully efficient firm. However, in reality you can observe only the firm's decisions that are suboptimal (inefficient) due to a variety of reasons. Stochastic frontier models, which are implemented in the QLIM procedure, allow for random shocks to production or cost along with technological or cost inefficiencies.
Time-varying-parameter unobserved component models
Unobserved components models (UCM) are a powerful and modern approach to modeling and forecasting time series data. These models are supported by the UCM procedure. The new RANDOMREG statement specifies regressors with time-varying coefficients. The new SPLINEREG statement models the effect of regressors that have a nonlinear impact on the dependent variable that is being forecast, and it can also model time-varying coefficients. These new features greatly extend the power of the UCM approach when independent variables affect the dynamic path of the dependent time series.
Zero-inflated count models
The COUNTREG procedure fits regression models for count data, where the dependent variable is a count that represents discrete events. Count data frequently display overdispersion and an excessive number of zero counts. Zero-inflated count models address these issues by assuming that the data are a mixture of two separate data generating processes, one for cases with a count of zero only, and one for cases with a count of zero or more. The COUNTREG procedure supports both the zero-inflated Poisson (ZIP) model and the zero-inflated negative binomial (ZINB) model.
Seasonal adjustment
The X12 procedure has many new statements and options and is now much more complete and up to date in its implementation of the current feature set of the U.S. Census X-12 program. In addition, in SAS/ETS 9.2 the X12 procedure now supports the TRAMO method for use in seasonal adjustment.
Numerous other new features
Important enhancements have been made in SAS 9.2 to many SAS/ETS procedures and access engines. These include support for copula methods for simulating multivariate distributions in the MODEL procedure and improved data access LIBNAME engine support for data supplied by the FAME data service, Haver Analytics, and the Center for Research in Security Prices.
Statistics and Operations Research Home Page | Econometrics and Time Series
