Use of Other Post-Hoc Tests with the Repeated-Measures Variable

In some instances, the contrasts described above will not be ideal, and you will need other multiple comparison tests for the interpretation of the analysis. For example, if there were ten levels of the repeated-measures factor, and if this factor were significant in the overall analysis, you might need to perform a multiple comparison procedure to determine which levels are different from each other. Although you could do this by using a large number of CONTRAST statements, this approach would be undesirable because each contrast is equivalent to a paired-samples t test. A multiple comparison procedure that controls for experiment-wise error rate would be preferable.

Several such tests are available in SAS/STAT for this situation (e.g., the Bonferroni test, the Scheffe test, the Tukey test). However, you cannot use these tests with the repeated-measures factor when you use the REPEATED statement. To utilize these tests, you must name the variable being analyzed in the MODEL statement. Fortunately, SAS can compute the ANOVA in a way that allows you to use these multiple comparison tests when analyzing a repeated factor. Essentially, this approach requires that the analysis be performed without use of the REPEATED statement. This procedure is somewhat more cumbersome than the approach previously described as it requires that the dataset be restructured. See Chapter 8 of Cody and Smith (1997) for ways to undertake such multiple comparison tests.