 |
 |
Hi, Not sure why you'd need afni, you can just load up the two zstat images in fslview and look at individual voxels to see if the statistics match (or are of opposite sign). The fixed effects model is simply a weighted average, where the weights are functions of the variances and so they are positive. That means the weighted average for a-b would be exactly -1* weighted average for b-a. As a double-check, I ran my own analysis of a-b and b-a at level 1 for 3 runs and then I ran the [1] and [-1] contrasts at level2 and the a-b for the [1] contrast exactly matches the b-a with the [-1] contrast. I viewed the zstats. I am using an older version of FSL, but I can't imagine it would matter. Not sure what's going on with your data. Generally, I only run the positive contrasts at the first level so I'm not eating up disc space. Plus the analysis runs faster since you're estimating half the contrasts. That said, the results should be identical to running both pos/neg at level one. Cheers, Jeanette On Wed, Jun 12, 2013 at 3:27 PM, Stéphane Jacobs <[log in to unmask]>wrote: > Jeanette, > > I've compared the average of cope1 (A-B) with the negative contrast of > cope2 (B-A) for completeness, and it's still the same problem... > > > Best, > > Stephane > > > Stéphane Jacobs - Chercheur post-doctorant / Post-doctoral researcher > > ImpAct - Inserm U1028 - Equipe Pélisson > Centre de Recherche en Neurosciences de Lyon > 16 avenue du Doyen Lépine > 69676 Bron Cedex, France > Téléphone / Phone: (+33) (0)4-72-91-34-20 > > Le 12.06.13 17:24, Jeanette Mumford a écrit : > > Hi, > > To be sure I understand your question, you are saying that at the second > level, when you are running a within-subject analysis to combine data over > multiple sessions for that subject, the average of A-B is not the same as > the negative of the average of the B-A contrasts? To be clear, say cope1 > is A-B and cope2 is B-A and the second level model, assuming 3 runs, is > 1 > 1 > 1 > > with the [1] and [-1] contrasts. You have data where the positive > contrast for cope 1 (A-B) at the second level is not equal to the negative > contrast for cope 2(B-A) at the second level? This doesn't seem > mathematically possible. Not that it should matter, but I assume you're > using fixed effects at the second level? > > Cheers, > Jeanette > > > > On Wed, Jun 12, 2013 at 4:46 AM, Stephane Jacobs < > [log in to unmask]> wrote: > >> Hello, >> >> I have what seems a trivial question but could not find an answer in the >> archives, so here it is... >> >> I have several runs per subject, each containing all of my conditions. I >> set up all the contrasts of interest at the first level, and then just do a >> cross-session average at the second level for each subject. In the >> contrasts I chose, I usually have a contrast and it inverse: A-B and B-A. >> As expected, at the first level tese give me inverse statistical maps. >> However, when I run the cross session average at the second level, this is >> not true any more, some supposedly inverse contrasts even giving sometimes >> very similar maps... >> >> Is there a reason why this should happen? Would it be better to approch >> this by averaging only the A-B contrasts at the second level, and defining >> the B-A contrast as its inverse at the second level (rather than average >> the first level B-A)? >> >> Thanks for any help or advice! >> >> Stephane >> > > >