 |
 |
Hi Mario > Dear SPM users, > > There have been questions similar to this posted in the list before, but I have doubts on a particular point I hope someone might clarify further. > > I have a group of subjects where I want to compare activation in response to tactile stimulation before and after a treatment. Thus I have two identical sessions per subject. In each session there is only one condition (stimulation) which is contrasted to baseline. The design is a simple block design with blocks of 32 seconds. > > I want to enter the two sessions in the same 1st level design for fixed effect analysis. When I do so I get a matrix where the first two columns represent a session each (see attachment). How then would I indicate when doing the contrast that I want to look at that effect (stimulation dependent activation) for that session? In other words > > (stimulation pre - baseline pre) - (stimulatin post - baseline post) a contrast 1 -1 will do note this assumes baseline is identical pre / post ; otherwise you have to make 4 columns activation/rest pre/post (although I don;t really like this kind of model since it is overspecified) and do [1 -1 -1 1] > > Is that what I am getting if I enter the vectors 1 -1 0 0 ? For a single session I would just use [1] as vector for that contrast (stimulation - baseline). In this case however, that within-session contrast is not specified. > > For the sake of clarification, let us say I wanted to look not at activation but deactivation during stimulation. For a sinlge session the vector would be just [-1], right?. How then would I compare deactivation between sessions? > > Would it make sense to enter the rest periods as another condition? In that case, what would the contrast vectors look like? for deactivation indeed -1 is the T contrast to use for comparing well it is a bit tricky - assume in a voxel that your values for pre/post are 10 and 5, it's cool the contrast [1 -1] gives 10-5 = 5 and 5>0 = more activated in pre than post - now if you have -10 and -5 then -10 -(-5) = -5 and -5<0 means more deactivated in pre than post ------> in SPM you need to run the opposite contrast [-1 1] that will be -(-10) + (-5) = 5 and 5>0 - now trickier you have -5 and -10, your contrast [1 -1] gives -5 -(-10) = +5 but what does it mean? not that pre is more activated but indeed less deactivated ... so to make things 'easy' you can do a F test [1 -1] which will look at all kind of situation and then plot the parameter estimate of each regressor (can be done by computing a F test [1 0 0 0; 0 1 0 0] then look at your F [1 -1] and do plot --> which contrast --> select the F [1 0 0 0; 0 1 0 0]) actually, only looking at the data will tell you what's going on (yes sometimes we need to look at the data not just rely on colored maps :-P ) Hope this help Cyril -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.