Check for outliers, equal variance, normality, and model validity

D1. Check covariate model: Examine SAS output

If your ANOVA model has a covariate, a regression variable used to explain differences among your experimental units, then an additional page of diagnostics are printed.

The purpose of the tests on this page is to verify that the standard covariate model is adequate for your data. The standard model has a single linear slope, applied to all treatment groups. This will not be appropriate if each treatment has a different slope, or if the slopes are not linear.

Therefore %MMAOV tests three terms:

• Covariate*Covariate*Treatment. This tests if the quadratic polynomial slope for the covariate differs among treatment groups.
• Covariate*Treatment. This tests if the linear slope for the covariate differs among treatment groups.
• Covariate*Covariate. This tests if a quadratic polynomial is needed, which can occur if the covariate has a curvilinear relationship with the dependent variable.

These tests are illustrated by the SAS output. Each section corresponds to one of the three tests, and you simply verify that ProbF is non-significant.

next >> ( Step D1 continued, Check covariate model:  Examine SAS output )