 |
 |
Hi, Just thought I would add my two pence worth. Classically, a difference between the contrasts is an interaction, and perhaps another way of looking for an interaction (i.e. a difference between differences) is to simply specify the contrast [+1 -1 -1 +1] and also [-1 +1 +1 -1], if your coditions are in the order [A1 A2 B1 B2] or [A1 B1 A2 B2]. As for looking for commonly activated voxels in the two contrasts, am I right in saying that Steves suggestion simply identifies voxels activated above threshold in both contrast, without specifically identifying those in which activity is significantly 'similar' (i.e. the degree of difference between A2>A1 and B2>B1 is statistically similar)? It would be nice to be able to do the latter, since this is the statistical inverse of an interaction. In other words, interactions are where the difference between differences is significant, and statistical conjunctions are where commonalities between the difference levels are statistically significant. Best wishes, Narender In message <[log in to unmask]> "FSL - FMRIB's Software Library" <[log in to unmask]> writes: > Hi. > > On Fri, 22 Mar 2002, Antti Korvenoja wrote: > > > I have a question concerning conjunction analyses. Or at least I believe it > > should be applied in my problem. If I have experimental conditions A1, A2 and > > B1,B2 and I want to determine what areas are significantly activated both in > > contrasts A2-A1, B2-B1, how do I set up this in FSL? > > I think you are asking - where is A2>A1 AND B2>B1 - yes? > > Then you want to take the cluster mask images from the A2-A1 (let's call > this contrast 1) contrast and the B2-B1 contrast (contrast 2) and > BINARY_AND (ie multiply) them together: > > avwmaths cluster_mask_zstat1 -mul cluster_mask_zstat2 -bin combined_mask > > this gives the desired binary mask image. > > > Or if my desire is to find > > areas that are differentially activated in these two contrasts? > > Do you mean (A2-A1)-(B2-B1)? that contrasts will give you where the first > contrast gives greater difference than the former. However - note that you > might well want to only look for differential activation like this where > both contrasts are also positive - in that case then mask this result with > the cluster masks (binarised to 0/1) from the individual contrasts. > > Hope this makes sense! ttfn, Steve. > > > Stephen M. Smith > Head of Image Analysis, FMRIB > > Oxford University Centre for Functional MRI of the Brain > John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK > +44 (0) 1865 222726 (fax 222717) > > [log in to unmask] http://www.fmrib.ox.ac.uk/~steve > -- Narender Ramnani FMRIB Centre University of Oxford