Fig. 8
GWA of whole brain fractional anisotropy data, using a whole brain parcellation of 42 regions (real data analysis). Permutation test was used to derive FWE corrected P-values of score test based on the simplified REML function using one-step random effect estimator. Among the 42 × 1,376,877 ≈ 57 million statistics, eight passed the permutation based FWE threshold (\(\chi _1^2 = 34.72\), blue line in Manhattan plot). Application of a Bonferroni correction for 42 tests to the usual GWA alpha level (5 × 10−8) yields to a more stringent threshold (\(\chi _1^2 = 36.98\), black line in Manhattan plot) where only one association survives, indicating the potential improved power from a permutation-based inference that accounts for dependency among the tests
