Rajesh Ranjan Nandy1
1Psychology and Biostatistics, University of California, Los Angeles, CA, USA
An important consideration fMRI is to choose the right threshold for activation. This is complicated by the temporal autocorrelation in fMRI data and the multiple testing involved in detecting activations. An ReML approach implemented in SPM2 to correct for the temporal autocorrelation but cannot eliminate the effects of inherent low frequency processes in resting brain. Also, the popular Gaussian Random Field approach to adjust for multiple comparison is usually not a vast improvement over the Bonferroni correction. We propose a novel approach using order statistics that adjusts for multiple comparison as well as the low frequency processes. No correction for temporal autocorrelation is necessary.