Hi,
The multiple comparison correction method that you use is arbitrary and you can use TFCE to avoid choosing a threshold (e.g. --T2). If you see an "empty" image, check that you have the display range set correctly (0.95 to 1.0 for significant results on the corrp image) and if you want to see if you have nearly significant results, but just under threshold, then change the display range (e.g. make it 0.8 to 1.0).
All the best,
Mark
On 26 Apr 2013, at 15:20, SUBSCRIBE FSL F. Nemmi <[log in to unmask]> wrote:
> Hello everybody!
> In a context of shape analysis following FIRST (two sample unpaired t-test), I recently start to use randomise.
> I have a concern about the threshold for the correction method stated in the subject of this mail: I chose to make t-contrast, and so I use the -S option. Anyway I have no clue (and actually I was no able to find any clue in the guide page or in the mailing list) about how chose the threshold.
> As a start, I left the threshold of "3" proposed in the example I found in the FIRST user guide page (although in the example the proposed correction method is different). Anyway, while the non corrected t contrast show some significant voxel, the corrected images are always "empty". I wonder if I simply have results which do not survive a multiple comparison correction or if, as I hope, I've committed a mistake in choosing the threshold (or the multiple comparison method).
> I hope someone has a hint for me
> Best regards
> F.
|