Integration of Data-Space and Statistics-Space Boundary-Based Test to Control the False Positive Rate

Many multivariate statistical methods have been applied to detect the difference between case and control population. However, it is difficult to control the false positive rate, especially under small sample size. Traditional family-wise error rate or false discovery rate adjusts the p values based on the distribution or ranks of p value in the same multiple testing. In this paper, we investigated the performance of integrating the Data-space boundary-based test (BBT) and Statistics-space BBT to control the false positive rate, under a previous proposed framework called Integrative Hypothesis Tests (IHT). The classification accuracy rate by Data-space BBT provides valuable information complementary to the p value from Statistics-space BBT. The simulation results demonstrated that the integration effectively controls the false positive rate even for small-sample-size cases. Experiments on the real-world dataset of bipolar disorder also validated the effectiveness of the integration.