Nilearn FDR correction: The FDR=0.05 threshold is inf

When using the nilearn threshold_stats_img function to perform FDR correction, the threshold returned is ‘inf’. Is this the correct output?

This means that no voxel is significant at FDR=.05.
Take a less stringent FDR control, e.g. FDR=0.2, if you want to obtain a finite threshold.

1 Like