Summer 2010 Workshop
Held on Thursday, June 3, 2010 at the
Crowne Plaza Hotel in Northbrook, IL.
The 26th Annual Summer Workshop
of the Northeastern Illinois Chapter of the American Statistical Association
will be given on the topic of:
SAS for Mixed Models
Walt Stroup, Ph.D., Professor & Dept. Head of Statistics, University of Nebraska
The course is intended for those who want to learn about the theory and application of generalized linear mixed models
from a non-Bayesian perspective. The material is presented at an applied level, accessible to participants with training
in linear statistical models and previous exposure to linear mixed models. Examples are drawn from a wide variety of allied disciplines.
We will begin with an overview of the main ideas of generalized linear mixed model theory. We make the connection between linear models,
generalized linear models, linear mixed models, and generalized linear mixed models (GLMM) in terms of model formulation, distributional
properties, and approaches to estimation and inference. The overview will include overarching issues that confront analysts who work with
correlated, non-normal data, such as overdispersion, the marginal and conditional models, and model diagnostics. Examples will use
SAS PROC GLIMMIX – accordingly, the introductory overview will include a brief look at GLIMMIX syntax.
The main focus of the examples will be on a variety of issues involved in mixed model analysis and its extension to rate and proportion
data and count data. The rate & proportion examples concentrate on binary, binomial and multinomial data in the presence of random
model effects and/or repeated measures. The count data includes various distributions (e.g. Poisson and negative binomial), their modeling
rationale and how to choose among them in practical situations. We also consider several zero-inflated models.
The final section focuses on power analysis and planning. GLMM theory allows comparison of competing designs not possible with conventional
power and sample size software. Most “conventional wisdom” about design assumes normally distributed data. This conventional wisdom is often
inappropriate – sometimes catastrophically so – for GLMMs. The bottom line is that the modeling, analysis, and design aspects of GLMMs cannot
be compartmentalized, but the design aspect has received relatively too little attention. This part of the course will illustrate tools that
can be used to help plan studies to effectively take advantage of GLMM capabilities.
Walter W. Stroup, Ph.D. is Professor and Head of the Department of Statistics at the University of
Nebraska, where he has been a faculty member since 1979. Dr. Stroup received a B.A. degree in psychology
from Antioch College, and M.S. and Ph.D. degrees in statistics from the University of Kentucky. Dr. Stroup
is coauthor of SAS for Mixed Models and SAS for Linear Models and he is widely published in statistical and
applied journals. He has presented numerous short courses on mixed models, nonlinear mixed models, and generalized
linear mixed models in academic, professional society, and industry based settings.