Clusterwise False Discovery Rate Control in Spatial Data

• Activity Number: 642
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Alexandra Chouldechova*+
• Companies: Stanford University
• Keywords: False discovery rate ; Spatial statistics ; Poisson clumping heuristc ; Multiple testing

Abstract

In numerous problems arising in epidemiology, genetics, and imaging, large scale multiple testing is carried out in order to identify regions of interest. A typical approach entails using a multiple testing procedure to test each location for the presence of signal. Contiguous clusters of rejected hypotheses are then reported. In such cases, it is often more informative to consider the clusters themselves to be the unit of inference. However, the location-wise error rate from the multiple testing procedure can greatly misrepresent the rate of erroneously reported clusters. It therefore becomes desirable to report the error rate on a cluster-wise rather than location-wise basis. The focus of this paper is to present a method for estimating and controlling the cluster-wise false discovery rate in such spatial data settings.