Cluster-level inferences

Nieto-Castanon, A. (2020)

Abstract

Whether in the context of voxel-based measures, such as when studying properties of SBC maps across multiple subjects, or in the context of ROI-to-ROI measures, such as when studying properties of RRC matrices across the same subjects, second-level analyses will perform a separate statistical test for each individual analysis unit (voxels in the former case or ROI-to-ROI connections in the latter). This often poses a considerable multiple-comparisons problem, where traditional false positive control strategies at the level of these individual units (e.g. thresholding individual voxels at a p<0.05 level) would result in unacceptably high rates of false positives across the entire analysis (e.g. across the entire brain, in voxel-based analyses). This chapter describes a number of strategies that have been developed to address this issue while simultaneously improving researchers' ability to make meaningful inferences from their second-level analysis results.