Dear Dr Anderson,
As always, I highly appreciate your detailed and direct responses! Thank you so much. Kindly, I have two follow-up questions about "FSL/randomise"
1. Is randomise correct for the number of contrasts?
2. In FSL wiki (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise/Theory) it was mentioned "though 10,000 permutations are required to reduce the margin-of-error to below 10% of the nominal alpha". Where the 10% came from? how it was calculated? e.g for 2000 permutations how can I calculate the margin of error?
Regards and thanks,
John
Hi John,
It's fine to use multiple contrasts. Only thing is that you may want to correct for them, i.e., multiple contrasts imply a multiple testing problem, that ideally should be accounted for. In PALM this can be done with "-corrcon".
All the best,
Anderson
On 23 February 2018 at 09:05, John anderson <[log in to unmask]> wrote:
Dear Dr Anderson,
I would like to inquire about GLM in randomise. For three groups (A, B and C) voxel wise analysis, I have three covariates and six contrasts). I understand that randomise takes the covariates into account in the voxel-wise analysis. Is this applies on the number of contrasts. I.e. is there any difference voxel-wise analysis with one contrast or six contrast.
Regards,
John
|