Are the analyses appropriate for answering the research questions?
Again, the research questions will dictate the analyses. If one is interested in relationships, a variety of correlational analyses are available, and the selection is influenced by the properties of the measures employed-nominal, ordinal, interval, ratio. And, one should not confuse a relationship with predictability. A correlation will provide the former, but multiple regression analysis is essential to determine the latter. If one is interested in making group comparisons, an analysis of variance may suffice. However, if groups differ on specific variables that may influence outcome, an analysis of covariance is preferred. Sometimes, we like our analyses so much we overindulge. Gluttony in statistical analysis can be eased by a Bonferroni procedure (Dunn, 1961), typically dividing the alpha (e.g., .05) by the number of analyses to control for experimentwise error, or a modified Bonferroni procedure (Larzelere & Mulaik, 1977). If a research report includes numerous analyses and the e
