Annotated Computer Output for Split Plot Design: SAS GLM.

Abstract : The analysis of covariance for split plot designs is not always straightforward when using a statistical software package such as SAS PROC GLM. In order to demonstrate correct analyses several data sets are examined and annotated SAS output is given. Hypothetical data are analyzed first without and then with the covariate included. The whole plots are arranged in a RCBD and the covariate is measured on the subplot experimental units. A second example has whole plots arranged in a CRD and the covariate measured only on the whole plot experimental units. Complete ANOVA tables for both examples may be computed in a single procedural call to SAS PROC GLM. Both Type I and Type III sums of squares are necessary to construct the proper ANOVA table. A commonly employed approach requiring two separate procedural calls to GLM is also demonstrated. Formulae for the standard errors of the difference between adjusted whole plot and subplot means are reported. This is part of a continuing project that procedures annotated computer output for the analysis of balanced split plot experiments with covariates. The complete project will involve processing three examples on SAS/GLM, BMDP/2V, SPSS-X/MANOVA, GENSTAT/ANOVA, and SYSTAT/MGLH.