Print

Print


Hi Desmond,

If you only want the positive direction of the one-sample t-test, you need to supply the "-t design.con" file containing that contrast (it's currently missing from your command line). The same file can contain the negative contrast).

If both design matrix (-d) and contrast (-t) are omitted, then it will automatically create design and both positive and negative contrasts for a one-sample t-test.

Also, unrelated: 10000 permutations with TFCE and NPC will be slow. Consider using fewer permutations and using the tail approximation. That is, use the options "-n 500 -approx tail -nouncorrected".

Also, consider using -logp, as the results are easier to read, with the figures generated lated having good contrast.

Hope this helps!

All the best,

Anderson


On 12 January 2017 at 21:29, Desmond Oathes <[log in to unmask]> wrote:
I think that I'm really close but maybe running the command incorrectly:

palm -i [modality#1] -i [modality#2] -d design.mat -n 10000 -T -npc -noniiclass -tonly -savemask -o OUTput

where design.mat is set to 1 for all subjs (EV1)
where design.con is set to '1' for contrast to do 1-SampleT (EV1)

> This doesn't do any sign flips/shuffling.  The tstat and npc 'fwep' maps just look like a binary brain mask.

Thank you,
Desmond