FLEXIBILITY AND POWER IN COMPARISONS AMONG MEANS

In general, psychological statistics texts advise the researcher to plan only orthogonal comparisons among means. This is done to avoid interpretation of dependent tests. It is argued here that if one uses prevalent post hoc techniques, the restriction of orthogonality in planned comparisons serves no useful purpose. A method is outlined in which comparisons among means are planned without the restriction of orthogonality. An index is provided which is a measure of the power of a set of k comparisons relative to that of a general post hoc technique. Use of the index allows one to trade off between number of comparisons and power while holding the Type I error rate constant. The result is increased flexibility relative to the planned, orthogonal approach and increased power relative to the post hoc approach. (PsycINFO Database Record (c) 2006 APA, all rights reserved). © 1969 American Psychological Association.