ABSTRACT
Merely performing multiple-comparison tests without a prior ANOVA can cause the same trouble as using many t tests. There is a plethora of multiple-comparison tests, generally named after the statisticians who developed them. Some tests also have a set of ranges (multiple-range tests) depending on how many means are between those being tested. Multiple-comparison tests are generally two-tailed so as to correspond to the two-tailed nature of the ANOVA. The multiple-comparison tests are performed at the same level of significance as the ANOVA. Multiple-range tests increase the range for some comparisons but not others. The aim is not to increase the range value for adjacent means because these are the closest and their differences are most likely to be missed. The procedures for the multiple-comparison tests are very similar. Instead of searching to find which means are significantly different, the means could be examined to show how they are related to each other: linearly or curvilinearly, and so on.
