Data Analysis Papers
Introducing the GLIMMIX Procedure for Generalized Linear
Mixed Models
Oliver Schabenberger, SAS Institute, 2005
Abstract
This paper describes a new SAS/STAT procedure for fitting models to non-normal or normal data with correlations or nonconstant variability. The GLIMMIX procedure is an add-on for the SAS/STAT product in SAS 9.1 on the Windows platform. PROC GLIMMIX extends the SAS mixed model tools in a number of ways. For example, it
- models data from non-Gaussian distributions
- implements low-rank smoothing based on mixed models
- provides new features for LS-means comparisons and display
- enables you to use SAS programming statements to compute model effects, or to define link and variance functions
- fits models to multivariate data in which observations do not all have the same distribution or link
Applications of the GLIMMIX procedure include estimating trends in disease rates, modeling counts or proportions over time in a clinical trial, predicting probability of occurrence in time series and spatial data, and joint modeling of correlated binary and continuous data. This paper describes generalized linear mixed models and how to use the GLIMMIX procedure for estimation, inference, and prediction.