False positives in neuroimaging genetics using voxel-based morphometry data

►RFT nonstationary cluster size test on VBM is found to be invalid based on empirical null studies. ►Invalid inferences (inflated false positive risk) were found with 3 voxel smoothing. ►With 6 voxel smoothing, invalid inferences were found with 6 voxel smoothing and alpha=0.01 cluster-forming threshold. ►RFT nonstationary test only found to be valid with 6 voxel smoothing with alpha=0.001 cluster-forming threshold. ►Equivalent Gaussian data simulations produced valid inferences, suggesting VBM data violates RFT assumptions and motivates the use of nonparametric permutation methods.