Multiple comparison procedures in analysis of covariance model
Abstract
In analysis of covariance, multiple comparison techniques are used to make inference about the treatment effects. An important feature of a good MC technique is controlling the familywise error rate for the family of tests considered. Except for the balanced designs, critical values of some simultaneous tests are not available. Simulation techniques are used effectively for the computation of the critical values for multiple comparisons. This investigation involves comparing the performance of the parametric and rank tests for multiple comparisons in the analysis of covariance models. The MC techniques considered are: Bonferroni, Scheffe’, Fisher’s stepwise least significant difference technique, and the techniques based on simulation. Robustness of these testing procedures is studied based on the violations of the underlying distributional assumptions. Under each study condition, performance of each test will be evaluated. SAS IML programming language is used for the simulation study.