Introduction to Generalized Linear Mixed Models

15 September 2013; 09:00 – 17:00

This full-day ENBIS-13 pre-conference course will be run by Prof. Tim Robinson (Department of Statistics, University of Wyoming)

The course is intended for anyone that would like an introduction to generalized linear mixed models.  Instruction will focus upon applications as well as developing an intuition on the use of these models. Interpretations of statistical output will also be a primary emphasis. Examples will be presented using the R statistical software package.  Participants are encouraged to bring along their personal laptops with the R software package pre-loaded.

Mixed models are a popular tool for analyzing correlated data due to the presence of random effects and repeated measures. In industrial applications this data often results from split-plot experimentation in which some factors have levels that are costly/difficult to change.  Repeated measures and longitudinal data require special attention because the inherent correlation must be addressed in order to obtain valid inferences. Normal theory models for repeated measures and random effects will be used for introducing examples of correlated data. These models will then be extended to generalized linear mixed models for and analysis of non-normal data included success/failure (binomial) and count data (Poisson).  Graphical analyses and residual diagnostics will also be presented.

The course is designed to encourage interaction among participants and with the presenter. No prior experience with mixed models is assumed.