Comparing one-sample t-test outputs from FSL randomise and FSL PALM

I have run a one-sample t-test in both randomise and PALM as a quality check.


randomise -i int_4D_all.nii.gz -o randomise_cope2 -T -1 -n 5000

palm -i int_4D_all.nii -o palm_cope2 -T -C 3.1 -n 5000 -save1-p 

PALM gives many more outputs compared to randomise when running the same test and wanted to clarify the corresponding outputs.

Questions arose concerning the processing and output:

  1. (a) Is it correct that palm_cope2_vox_tstat_c1 is equivalent to the output randomise_cope2_tstat1 (which represent the raw tstat)? (b) Additionally, would it also be the case that palm_cope2_tfce_fwep_tstat_c1 is comparable to randomise_cope2_tfce_corrp_tstat1 (which represents the corrected p-values for tfce thresholding)?
  • Based on my visual comparison, the values at various voxels align for (a) and are similar for (b) but not exact (e.g., 0.337 for randomise_corrp vs. 0.284 for palm_tfce_fwep). This would be presumable given that it underwent permutation testing, correct?
  1. PALM was exponentially slower than randomise (4 hours vs. 15 minutes). Why is this the case when the inputs and tests conducted are the same?