Group Sequential Robust Designs in Genetic Studies Open Access
Downloadable ContentDownload PDF
The statistical analysis methods of the data collected in genetic studies are locally optimal under the specified genetic model. Applying a statistical test that misspecifies the genetic model can result in significant loss of power. To accommodate this problem, the robust test statistic MAX, the maximum of all test statistics, is preferred. However, it no longer follows asymptotic normal distribution. This dissertation presents approaches that combine the group sequential design and the robust approaches of statistical testing in various genetic study designs. It describes methods of obtaining critical values for setting up boundaries in a two-stage group sequential design using MAX in order to control for overall type I error. Optimal designs varying by different designs and genetic parameters are studied. Finally, this dissertation also presents adaptive designs by two proposed genetic model selection methods as alternative robust methods to prevent significant power loss due to model mis-specification.