Dear list,
I try to make my question a little clearer. I performed tbss across subjects belonging to three independent groups. I then used randomise the compare them with the model attached. I like to ask now whether I can consider p<0.05 as significant for each of the 6 comparisons or whether I need to adjust the statistical threshold somehow for multiple testing (i.e. the six comparisons created in the model)?
Thank you very much for your suggestions.
Best wishes
André
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]]" im Auftrag von "André Schmidt [[log in to unmask]]
Gesendet: Dienstag, 18. November 2014 17:07
An: [log in to unmask]
Betreff: [FSL] Correction for multiple comparisons
Dear list,
I used randomise to compare FA values obtained from tbss across 3 groups - in the design matrix I created 6 contrasts (a-b,b-a,a-c,c-a,b-c and c-b).
Now i am wondering whether i have to adjust the p value for multiple testing (6 contrasts). I also see that randomise corresponds to a 1-sided t test but i guess i have to test 2-sided, so do i need to consider p=0.05/(6x2) as significant or is this too much correction?
Thanks a lot for your help.
Best wishes
André
|