Both the concern about selective inference and the concern about the right level of variability are recognized and addressed in the ongoing efforts to map the functioning brain (via fMRI).

Our research contributes to the first concern by offering a new approach to Voodoo Correlations, a term used for higher than realistic correlations that emerge once selected from all over the brain. So far these were addressed by offering appropriate confidence intervals, to be followed by estimators. We also develop hierarchical methods to address the effect of selection when the interaction between brain and genes are at discovery arena.

It has now become common to take a wider view of variability by using a mixed model analysis for activity in a brain location where the effect of individual subjects is treated as random. Instead we recognize that there may be an effect in a location for some subjects, and none for others. We therefor suggest prevalence maps as a way to capture the right level of uncertainty in describing functional activity.