All Times EDT
Keywords: Mean test; Median test, Noninferiority test, Equivalence test; Bootstrapping; Permutation
When comparing the means of normally distributed endpoint of two drug products, t-test is used as a standard procedure. When the data fail the normality assumption, but the sample sizes are sufficiently large, a large sample z-test is used. However, when the sample size is small, the conventional non-parametric approaches such as median test, tests based on the rank or tests based on transformed data that are applicable to test for superiority may not be able to be applicable to non-inferiority or equivalence hypothesis with a non-zero margin set for mean difference. In this presentation, we discuss the difficulties of applying median and rank test for such hypothesis. We further discuss the potential application of bootstrapping and permutation tests.