To be or not to be: Control and balancing of type I and type II errors

The various costs of Type I and Type II errors of inference from data are discussed. Unfortunately, other things equal, Types I and II errors are inversely related. Six methods of minimizing Type I error in studies are presented, each of which may be employed even after data are collected. Type II errors may also be minimized by a combination of study design and analytic means. Although some of these techniques must be built into the study others can be undertaken when beginning work on a data set. Only one of these techniques will increase Type I error, and those involving data reduction may decrease both types of error.