Fig. 6

Comparison of test statistic values for association testing (H0:β1 = 0) using real data analysis. Results shown for the non-iterative and fully converged random effect estimators based on the simplified REML function and FaST-LMM’s LRT. Each plot represents a ROI where x-axis shows FaST-LMM’s LRT and y-axis represents the score test. Despite strong concordance between the score test results using WLS or fully converged random effect estimator, FaST-LMM is slightly more powerful, consistently for whole brain parcellation (see Supplementary Figures 10 and 11 for the rest of the ROIs)