ABSTRACT
This chapter summarizes the major considerations that should be taken into account when choosing the best statistical technique. I have presented in this book a variety of statistical techniques related to categorical and nonparametric data analysis, techniques that can be used as alternatives to traditional parametric technique (e.g., ANOVA, ordinary least squares regression). In choosing the best statistical technique for a given situation, one needs to consider two major questions: (a) Is use of the technique valid in the situation; that is, are all assumptions met? and (b) Is the technique more statistically powerful than other ones? In my opinion, the first consideration is a bit more important than the second, because it makes little sense to maximize power if the resulting estimates of p values are biased or inconsistent. (I grant that a researcher may sometimes find it permissible to allow a minor violation of assumptions and a small amount of bias for the sake of power, but the burden of proof should be on the researcher to justify such a trade-off.) It should be noted that violation of assumptions can increase Type II errors as well as Type I errors, so considerations of validity and power are not always contrary.
