Handling Multiple Testing in Local Statistics of Spatial Association by Controlling the False Discovery Rate: A Comparative Analysis

Handling multiple testing plays a key role in spatial pattern detection by using local statistics of spatial association. In recent years, it was found that the FDR (False Discovery Rate)-based corrections are more powerful than the FWER (Family-wise Error Rate)-based corrections. Although different FDR-based corrections have been proposed, a systematic and comprehensive comparison of different FDR-based corrections is noticeably absent. In this study, a comparative study of nine FDR-based corrections for handling multiple testing in local statistical of spatial association is performed. Experimental results show that there are significant differences among different corrections. Specifically, corrections under the assumption of independence are remarkably powerful than those under dependence. Of all the corrections the assumption of independence, the GBS is more powerful than others, however, no significant difference is found among them. It is also found that all the FDR-based corrections are still conservative.