Comparison of Features and Applications of four Linear Models Procedures

Development of SAS ® linear models procedures over the past several years has led to a number of easily accessible methodological statistical advances for experimental data analysis. The original linear models program, GLM, was a fixed model procedure for analysis of normally distributed data with homogeneous variances. The GENMOD procedure extended the fixed linear model analysis to a number of non-normal distributions. With the use of GEE, GENMOD was able to address correlated repeated measures data. The MIXED procedure permitted the user to model both fixed and random effects for normally distributed variables. Because modeling of random effects permits multiple residual error terms, it is frequently possible to model heterogeneous residual variances. The most recent linear models procedure, GLIMMIX, has the capabilities of GLM, GENMOD and MIXED in one procedure. This presentation looks at the unique features and appropriate applications of each of these linear model procedures.