ABSTRACT
This chapter presents several multiple comparison procedures. It presents the most commonly used multiple comparison procedures in the setting of a normal completely randomized design model with the knowledge that they can also be used in analyzing data from other models to be studied later. The chapter considers two procedures that are based on the t–distribution: the Bonferroni multiple comparison procedure and Fisher's "protected" least significant difference procedure. It illustrates the usage of Statistical Analysis System for multiple comparisons with the pain threshold data. Several competing multiple comparison procedures have been developed, based on different properties. Some procedures provide both simultaneous confidence intervals and tests, while others provide only tests. Some procedures apply directly to equal sample size problems, while others apply to both equal and unequal sample size problems. In the multiple comparison procedures presented thus far the same critical difference is used in making all comparisons.
