hi everyone
I am posting to this thread because I have basically the same questions. I know Bob has answered similar questions elsewhere but I cannot find answers to these particular ones.
Quote
Why are there MORE voxels required to meed 1-sided thresholding here?
I think this has to do with the difference between the terms 'one-sided' and 'one-tailed' as it applies to using 3dClustSim and then thresholding in the AFNI viewer and using Clusterize. An explanation of this would go a long way for me, 3dFoodie and maybe(?) a few other people as well. As Bob notes, SPM does one-sided thresholding but it appears to also do one-tailed tests, so that the critical t value it gives for, say, p=0.001 is the critical t for p = 0.002 in a two-tailed test with the same df and therefore the cluster size thresholds are more liberal.
My questions, drawn from the above posts, are:
1) When clustering an F-map resulting from an ANOVA, which value to use? My instinct says the one-sided, since all Fs are nonnegative.
2) If I am interested in both positive and negative effects in a t-test, and I want to carry it out such that my overall thresholding reflects a two-tailed test at say p < 0.001 (in other words, that I am putting 0.0005 into each tail), then which type of thresholding and what critical t-value should I use?
I'm trying to be as clear as I can, and I hope these questions seem reasonable.
Quote
We give you lots of choices -- the price of which is that you have to understand what is going on in order to make a correct decision.
A corollary of which is maybe that the more choices you have the less you feel like you understand what is going on (always a good thing!).
James