Nonparametric vs parametric tests of location in biomedical research.

d t p d t A d n HE CHOICE OF STATISTICAL TEST HAS A PROFOUND impact on the interpretation of data. Understanding this choice is important for the critical evaluation of he biomedical literature. The question often arises hether to use nonparametric or parametric tests. The test is the most widely used statistical test for comparing he means of 2 independent groups. This parametric test ssumes that the data are distributed normally, that samles from different groups are independent, and that the ariances between the groups are equal. The most comonly used nonparametric test in this situation is the ilcoxon rank-sum test (WRST) and the closely related ann–Whitney U test. The WRST assumes that observaions from the different groups are random samples (ie, ndependent and identically distributed) from their respecive populations and are mutually independent and that he observations are ordinal or continuous measurements. hen there are k groups (treatments), the nonparametric est is the Kruskal–Wallis test (KW), a generalization of he WRST. KW is the nonparametric equivalent to nalysis of variance (ANOVA). Using nonparametric tests nstead of parametric tests brings about 2 questions: 1) hat happens if the nonparametric test is used when the arametric assumptions are met?; and 2) What happens hen the parametric assumptions are not met? To answer these questions, one must first discuss the nderlying goal of the study. Usually in biomedical appliations one is interested in measures of location such as the ean. One can test if the treatment (experimental condiion) has an effect (location shift) on the population under tudy. For example, one may be interested in the effect of reatment(s) on a specific measurement, say cell count, ompared to the control. Data of this nature are often nalyzed with the t test, or if there are k 2 groups, NOVA. In the parametric case, one tests for differences n the means among the groups. In the nonparametric case, quivalents the location statistic is the median. The assumptions for the nonparametric test are weaker han those for the parametric test, and it has been stated