New in SAS/STAT 9.3 is an experimental procedure for fitting finite mixture models.  The FMM procedure fits statistical models to data for which the distribution of the response is a finite mixture of univariate distributions.  PROC FMM can model both the component distributions and the mixing probabilities. You use finite mixture models for applications such as estimating multimodal densities, modeling overdispersed data,and fitting regression models with complex error distributions. See Gentle Introduction to Finite Mixture Models for more information. 

Forecasters often deal with data accumulated at different time intervals (for example, monthly data and daily data). A common practice is to generate the forecasts at the two time intervals independently so as to choose the best model for each series. That practice can result in forecasts that do not agree. This paper shows how the SAS High-Performance Forecasting HPFTEMPRECON procedure uses the lower-frequency forecast as a benchmark to adjust the higher-frequency forecast to take the best advantage of both forecasts.

The shared frailty model is a useful technique when you face correlated time-to-event data. It incorporates cluster effects into the model as random variables. The PHREG procedure in SAS/STAT 9.3 provides a RANDOM statement which enables you to fit a shared frailty model.  This example illustrates that technique and also demonstrates the marginal Cox approach, which uses a robust sandwich covariance estimator to account for the intracluster dependence.

This new Web example shows you how to use the SURVEYMEANS procedure and a little SAS programming to estimate the finite population standard deviation when you have data from a complex survey design.

SAS/OR 9.3 adds a variety of new and enhanced operations research capabilities. This paper surveys these added capabilities and highlights their advantages and benefits. In optimization, new nonlinear and quadratic solvers provide improved performance, and a multistart algorithm for nonlinear solvers enables you to search more effectively for globally optimal solutions to difficult problems. SAS® Simulation Studio adds support for 64-bit Windows environments and enhances graphical plots and icons, model editing and viewing, and distribution fitting with JMP® software.  The CLP procedure enhances its activity-scheduling capabilities and adds support for a linear objective function, providing an alternate means of optimization.

Just starting to wade into Bayesian waters? This updated resource page for using SAS/STAT software for Bayesian analysis might be just the ticket.  It includes useful links and information about our dual approach to Bayesian methods: putting facilities for key statistical analyses into existing procedures and providing a general purpose Bayesian modeling tool.

We frequently receive requests from users who want some help in presenting the highlights of a new release to a work group.  The new Ambassador Kit includes the slides used in the What's New in SAS/STAT® 9.3 video, complemented with notes and a resource paper.  Download this kit and let us know if it’s useful.

Our video portal grows with the addition of three new videos:  New Features in SAS/OR® 9.3, New Features in SAS® Simulation Studio and Calling R from SAS/IML® Software.  Keep checking back on our video site because we have even more videos under production.  We hope that these videos provide a quick and easy overview of our latest analytical capabilities for you.   They are also available on the SAS YouTube channel.

Rick Wicklin provides a great explanation of how random number streams work in SAS in his statistical programming blog. Read this post to find out more about using the RAND and RANDGEN functions and how to create a reproducible stream of random numbers both in the DATA step and in the IML procedure.

SAS and JMP will exhibit jointly at the 2011 Winter Simulation Conference in Phoenix from Dec. 11 to Dec. 14. Emily Lada and Ed Hughes will present "Introduction to SAS Simulation Studio," and Brady Brady and Ed Hughes will present "Experimentation, Exploration, and Simulation with JMP and SAS Simulation Studio."

PROC OPTMODEL does not provide a BY statement to obtain separate analyses for groups of observations defined by one or more grouping variables. This example illustrates how BY group processing can be done within PROC OPTMODEL using a DO loop.