Why Does the Permutation Code Run Faster when Using a Small ROI?
When using an ROI to constrain the search, hundreds of thousands of permutations should run very fast (e.g. half an hour) as they only count significant effects that occur in a small region of a surface. Permutation Tests – Separating Positive and Negative Effects (programs in $ROI): arb_surfNstat4D_Pmap_randomize10000_ROI_POS_NEG This new permutation code separates out positive and negative effects of a covariate. It first (as before) gives a p-value for the overall effect of a covariate (such as group membership, age, SAPS, IQ, etc.). Then it also computes a p-value for the positive effects, and a separate p-value for the negative effects. It reports all 3. The neat thing is the p values are often better than before, often by a factor of 0.5, as it is a bit like doing a ‘one-tailed’ permutation test in a case where there is a one-sided hypothesis. An example script (based on Kevin’s surfaces) is in: /data/ad/mass3/users/PAULS_SURFACE_CODE/SGI/4D_UCFS/ARB/ROI/testy How it works. It wo
