Experimental Statistics

SAS Software Logo
JMP Software Logo
R Software Logo
IBM SPSS Software Logo

The Experimental Statistics Unit provides statistical support to researchers on the University of Georgia Griffin Campus. Below you will find previous seminars which can serve as resources that may help you along your research data analysis.

Linear Mixed Models with Random Effects

Analysis of variance models that include random effects are called linear mixed models with random effects (LMM). For agricultural experiments, random effects are often reps, blocks, years, locations or subjects. Usually, random effects are not tested, but are used to adjust the standard errors for tests on the fixed effects. This broadens the inference by implying that one would expect the same results regardless of the block, location or year in which the experiment was conducted.

Linear Mixed Models with Repeated Effects

Repeated effects are factors where the experiment unit is measured repeatedly. Common factors are time or spatial increments, such as dates or soil depth.  The assumption of equal variance within each time or space measurement is no longer necessary because the variances can be estimated by the model. A repeated measures model can also include random effects.

Generalized Linear Mixed Models for Count Data

Generalized linear mixed models (GLMM) allow for the modeling of non-normal response data within an analysis of variance framework. The distribution of the response data is included in these models so the assumption of normality is unnecessary. Count data usually has a Poisson or negative binomial distribution and was the focus of this seminar. Binomial and multinomial are two other types of response data often seen in agriculture experiments. They can also be modeled  with a generalized linear mixed model.

Seminar References

Several of the most useful papers and one book listed in the reference sections of the seminar handouts.

Tips and Strategies for Mixed Modeling with SAS/STAT® Procedures

A popular paper given at several SAS® Global Forum conferences. To paraphrase, this paper has recommendations for improving performance, methods for obtaining convergence and explanations of various notes, warnings, or error messages. The authors work in SAS Technical Support and see lots of questions and problems.

Analysis of Generalized Linear Mixed Models in the Agricultural and Natural Resources Sciences

This is arguably the best reference book for fitting generalized linear mixed models to data from agricultural experiments. While it does include some complex equations and theoretical concepts of generalized linear models, the accompanying text is geared toward applied researchers and analysts. Practical examples from the agricultural sciences accompany each type of model that is discussed.

Making Comparisons Fair: How LS-Means Unify the Analysis of Linear Models

Everything you wanted to know about Least Squares Means (and probably a little bit more).

Rethinking the Analysis of Non-normal Data in Plant and Soil Science

An Agronomy Journal paper in which the author compares modeling data the old way using transformations and PROC GLM to modern methods using generalized linear mixed models with PROC GLIMMIX.

Living with Generalized Linear Mixed Models

An earlier version of the previously mentioned paper that introduces generalized linear mixed models and discusses why they are better than traditional modeling strategies.