In the Fit Least Squares report, use the Multiple Comparisons option to obtain tests and confidence levels that compare means defined by levels of your model effects. The goal of multiple comparisons methods is to determine whether group means differ, while controlling the probability of reaching an incorrect conclusion. The Multiple Comparisons option enables you to compare group means with the overall average (analysis of means) and with a control group mean. You can also conduct pairwise comparisons using either Tukey HSD or Student’s t. You can also perform equivalence tests to identify pairwise differences that are of practical importance.
The Student’s t and equivalence testing methods control only the error rate for an individual comparison. As such, they are not true multiple comparison procedures. All other methods provided control the overall error rate for all comparisons of interest. Each of these methods uses a multiple comparison adjustment in calculating p-values and confidence limits.
If your model contains nominal and ordinal effects, you can conduct comparisons using Least Squares Means estimates, or you can define specific comparisons using User-Defined Estimates. If your model contains only continuous effects, you can compare means using User-Defined Estimates.
Tip: Suppose that a continuous effect consists of relatively few levels. If you are interested in comparisons using Least Squares Means Estimates, consider assigning an ordinal (or nominal) modeling type to that effect.
This section contains information about the following topics:
• Launch the Multiple Comparisons Option