Chapter 6. Nonparametric Methods in Pharmaceutical Statistics

Paul Juneau^[]

^[] Paul Juneau is Associate Director, Nonclinical Statistics, Pfizer, USA.

Nonparametric, or distribution-free, statistical methods are very useful in the setting of pharmaceutical research. These methods afford data analysts the ability to relax some of the assumptions typically made by their Gaussian (normality-based) analogues. In some settings (e.g., drug discovery investigations), these assumptions may not be verifiable due to small sample sizes. In others, where larger sample sizes are employed (e.g., clinical trial settings), the assumption of a Gaussian (normal) distribution is not met because of the presence of heavy-tails in measurement response or a large degree of skewness. This chapter covers two settings found commonly in pharmaceutical research (two-sample setting and one-way layout) and discusses sample size determination in a nonparametric sense. The introduced statistical methods are illustrated using examples from drug discovery studies and clinical trials.