Researchers select a multiple comparison procedure on the same bases on
which any other statistical procedure is selected, that is, the best control of Type
I and Type II errors. Researchers select the hypothesis testing procedure that min
imizes the probability of making a Type II error (or equivalently, one that maxi
mizes the statistical power) while maintaining the probability of a Type I error at
or below the nominal alpha. The issues of probabilities of both types of errors are
somewhat more complicated in the context of multiple comparisons than in the
context of a single hypothesis test