Multiple comparisons involve the situation when more than one statistical test of significance is conducted on a single set of data (Hu 1999). Statistical tests estimate the probability that the observed difference could occur on the basis of chance findings. If p < 0.05, the difference found is likely to be seen only five times if the experiment is repeated 100 times. If more than one comparison is completed, the experiment-wide error rate changes. In these instances, the probability level is inflated, and the researcher may wish to control for experiment-wide error with a mathematical correction. For example, a study may investigate the effect of age on memory, abstract problem-solving, verbal fluency, spatial analysis, and attention. Having multiple comparisons raises the probability that a spurious significant difference may be found for one or more of the variables.
For a priori comparisons, that is, for comparisons that are planned on the basis of...