Parametric Analysis of Ordinal Categorical Clinical Outcome

Parametric and nonparametric approaches are compared for testing hypotheses of the superiority of one therapy over another when the outcome is ordinal categorical. The estimation of clinically informative effect sizes, especially the probability that a client given one therapy will have an outcome that is superior to that of a client given another therapy, is emphasized. Parametric and nonparametric estimates of effect size are presented. Considerations for choosing between the two approaches are discussed.