Solving the multiple comparison problem in fMRI with a novel nonparametric approach using bootstrap in autoregression

R. R. Nandy, D. Cordes University of Washington, Seattle, Wa, United States Synopsis The multiple comparison problem has always been a challenging one in fMRI due to the complex nature of spatial dependence among neighboring voxels. A popular conservative solution is the Bonferroni correction which works well when the hypotheses are independent but turns out to be too conservative for fMRI analysis. A better approach is the Gaussian random field approach, but it makes several strong assumptions, the validity of which cannot always be justified. Here we propose to solve the multiple comparison problem by bootstrapping resting state data while preserving its autoregressive structure and then calculating the distribution of the maximum statistic.