Dear Anderson, dear all
I did some test, but with no success.
i explored both approaches:
1) manually subtract 1st level analysis
- I duplicated all the 1st level folder of session 2 creating a feat_folder_diff
- i subtracted the cope1 and the varcope1 of session2 with those of session1, in the standard space (feat_folder/reg_standard/stats/copeX)
$FSLDIR/bin/fslmaths $sess2dirstd/cope1 -sub $sess1dirstd/cope1 $sessDiffdirstd/cope1
- I transformed the cope difference image from standard to individual and copied to : feat_folder_diff/stats folder.
- I did a group analysis that seems erroneous.
the FEAT Time Series Report - zstatX are empty...no graphs, full_model, partial_model, raw_data are all 0s.
did I make some errors??
2) create a 2x2 model mixed model
I admit my problems in understanding the proper model to be used. I mined the list and the fsl wiky but I could not find the right solution.
the Multi-Session & Multi-Subject (Repeated Measures - Three Level Analysis) create a mean of the 3 sessions, while I'm interested in the difference, so it's no worthy.
assuming 4 subject with 2 sessions, does the following model estimate a two sessions difference??
1 0 0 0
-1 0 0 0
0 1 0 0
0 -1 0 0
0 0 1 0
0 0 -1 0
0 0 0 1
0 0 0 -1
with the contasts
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
if it were correct, that would be my favorite solution, as I would get one feat folder for each subject, so using many 3rd level analysis I could test group comparisons with several covariates/nuisance schema.
in a previous post (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;20f08063.1407) I found a 2x3 model which contain most of the contrast I would like to do.
It uses 2 columns for each group (i refer to ev1-2 & ev3-4). I assume for there are 3 levels. with 2 levels how does it reduce to that model ???
thanks in advance
cheers
Alberto
|