Resampling based confidence intervals in flexible sequential designs

For controlled clinical trials sequential designs have been proposed which e.g. allow for mid-trial adaptation of sample sizes and test statistics to be applied after each interim analysis without inflating the type I error probability. Although important for applications, confidence intervals following such flexible sequential trials have been proposed only recently. All suggested confidence intervals assume normal distributions for the stage wise test statistics, an assumption usually justified by asymptotic arguments. However, normal distributions might not be achieved with sufficient accuracy if in a sequential design the stage wise sample sizes are too small. For single stage designs often the Bootstrap or an other resampling approach leads to a more accurate method. It is the aim of this research project to develop resampling methods for confidence intervals following a flexible sequential trial, and to evaluate their performance.