Hi Anderson,
Thanks for the follow up. Your last example is how I started to set everything up initially so this makes more sense to me as I am not a stats expert by any means. Based on your feedback I fixed the design.con and design.mat files. The outputs (when I ran a short simulation of 50) where in .mgz format to match the inputs.
1)When I used mri_convert to change it to .nii the files were still not recognizable by fslview or other viewers. What is the proper way to convert these images .mgz stats files to NIFTI?
For the analysis I am currently running the following:
palm -i lh.thickness.10mm.mgz -i rh.thickness.10mm.mgz -d design.mat -t design.con -o thickness -n 500 -approx tail -corrcon -corrmod -s lh.white -s rh.white -T -logp -m lh.mask.mgh -m rh.mask.mgh -twotail -save1-p
2) In the approx tail option does it use the "heavy tail" approximation which came out recently or does it assume a gaussian distribution?
3) For the mask files I ran qdec on left and right hemisphere itself and used the mask files created there. Does palm automatically merge the left and right thickness files and masks into one file or do I run that step manually first? Would leaving it separate make much of a difference?
4) On other stats packages I typically run a 2 tail tests, and so I enabled this option since we used -corrcon and -corrmod instead of the npc versions. Should I really be using npc versions? Is this assumption of adding 2-tail correct?
Thanks again for all of the help...It really helped to get this up and running!
Best,
Ajay
|