ABSTRACT
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
Due to the large size of neuroimaging data, the question of corrections for multiple testing cannot be ignored if valid inference is to be obtained. Historically, familywise error rate has been a relevant quantity to control; for massive datasets, however, this criterion is generally too conservative. More recently, methods for the control of the false discovery rate have gained in popularity in the functional neuroimaging literature, as they are better able to accommodate a large number of tests carried out simultaneously. Bayesian procedures and approaches that explicitly account for the spatial correlation in brain image data are also becoming more prevalent.
