Title:A studentized permutation test in group sequential designs

Abstract:In group sequential designs, where several data looks are conducted for early stopping, we generally assume the vector of test statistics from the sequential analyses follows (at least approximately or asymptotially) a multivariate normal distribution. However, it is well-known that test statistics for which an asymptotic distribution is derived may suffer from poor small sample approximation. This might become even worse with an increasing number of data looks. The aim of this paper is to improve the small sample behaviour of group sequential designs while maintaining the same asymptotic properties as classical group sequential designs. This improvement is achieved through the application of a modified permutation test. In particular, this paper shows that the permutation distribution approximates the distribution of the test statistics not only under the null hypothesis but also under the alternative hypothesis, resulting in an asymptotically valid permutation test. An extensive simulation study shows that the proposed permutation test better controls the Type I error rate than its competitors in the case of small sample sizes.