Hello,
We have been experimenting with Threshold-free Cluster Enhancement using the new pTFCE toolbox in SPM as well as the existing implementation in FSL’s randomise function. We want our final result to be corrected for family-wise error at the .05 threshold, and believe we have correctly followed the instructions for each tool, but are surprised by the large differences in results between the two, so we wonder if we might be doing something wrong. With pTFCE in SPM, we are thresholding the “enhanced” contrast image at FWE <.05, and with randomise we're thresholding the respective “_tfce_corrp_tstat” (1-P) image at .95, and we are finding that the randomise output seems much more liberal and in fact more closely resembles the pTFCE image thresholded at uncorrected p<.05.
We’d love any thoughts that might help us understand these differences, and would be happy to provide more information.
Thanks!
Annchen Knodt
|