Creation date: 05/28/98
Authored by: Rich Herrington

Question:

What are the differences between SAS's PROC MIXED and PROC GLM?

Answer:

GLM is an acronym which stands for "General Linear Model". PROC GLM allows one to fit both fixed effect regression models and fixed effect ANOVA models. All coding for categorical predictors and interaction terms are conveniently created by SAS internally. PROC GLM offers several algorithms for calculating "sums of squares" (Type I to IV SS). PROC MIXED subsumes PROC GLM as it allows testing of both fixed effects and variance(covariance) components. PROC MIXED uses iterative optimization methods (newton-raphson method) that maximize a likelihood function. As such, there are no sums of squares calculations. PROC MIXED allows for fitting of quite complex mixed-effect linear models such as random coefficient models and hierarchical linear models. PROC GLM cannot accommodate model these kinds of models.

BACK | MAIN

Last updated: 01/18/06 by Karl Ho