Fig. 3.

Averaged results based on 10.000 permutations in which random samples were drawn with sample size equal to the smallest dataset (pure migraine, N = 335 cases and 1155 controls). Results are shown for predictions based on RADIANT (Fig. 3a), ATM (Fig 3b), and ATM after exclusion of 805 MDD cases (Fig. 3c). Note that for pure migraine, the original results are shown since no permutations were run for the smallest dataset