Fig. 13.

Empirical false positive rate (red) obtained from permutation testing. We used 100 random permutations of a subset of 60 normal subjects with the original design and subsequently inverting the model including 300 000 gray matter voxels. Posterior probabilities were thresholded using various thresholds (e.g. 0.95) and false positive rate was estimated as the number of above threshold voxels per volume averaged over the all permutations. The histogram of obtained false positive rate is shown right.