andersonwinkler / PALM

PALM: Permutation Analysis of Linear Models
69 stars 28 forks source link

P-value outputs are empty for single group average #33

Open btaraku opened 3 years ago

btaraku commented 3 years ago

Dear Anderson,

I am running a single group average on my fMRI dataset, and I am having trouble getting all of the outputs that I am looking for. I am able to get activation in the tstat, cluster extent, and TFCE outputs, however, the uncorrected, FWE corrected , and FDR corrected p-value outputs for each of the 3 statistics mentioned are all empty (no activation across the brain). I am particularly interested in getting my TFCE and cluster extent outputs in a more interpretable form, such as a zscore, but I can only get the default values for these stats. Do you have any suggestions of how to convert the cluster extent and TFCE values to p-values?

I am including the palm commands I am using in my bash script to generate these stats. The first line runs palm with cifti input, and the next two lines run palm with nifti and gifti inputs (separated from the cifti) in order to run cluster extent and TFCE.

Run palm without tfce and clustere

palm -i Y.${extension} -transposedata -d ${workdir}/design.mat -t ${workdir}/design.con -o results_cifti -n ${niter} -corrcon -logp -accel tail -zstat -fdr

Run palm with tfce and clustere

palm -i Y_subcortical.nii -d ${workdir}/design.mat -t ${workdir}/design.con -o results_dense_subcortical -n 500 -corrcon -C ${cluster_thr} -Cstat "extent" -logp -accel tail -T -zstat -fdr

palm -i Y_left.func.gii -i Y_right.func.gii -d ${workdir}/design.mat -t ${workdir}/design.con -o results_dense_cortical -n 500 -corrcon -corrmod -C ${cluster_thr} -Cstat "extent" -logp -accel tail -T -tfce2D -s L_midthickness.surf.gii -s R_midthickness.surf.gii -zstat -fdr

Please let me know if you need me to clarify anything, your help is greatly appreciated.

Thank you, Brandon

andersonwinkler commented 3 years ago

Hi Brandon,

Usually the uncp does show non-zero results. I wonder if this isn't a masking issue of sorts. Could you share your input files? If so, send a link by email (use my username here on github at gmail).

Cheers,

Anderson