Dear everyone,
In case of FIR predictors and overlapping trials within a condition, for an illustrative example e.g. think of a condition with two onsets aligning to the 1st and the 5th TR, window length 5 TRs, order 5,
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
1 0 0 0 1
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 0 1
model specification results in a design matrix with a pattern
1 0 0 0 -0.5
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
1 0 0 0 0.5
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 0 1
This is due to the orthogonalisation in spm_fMRI_design lines 242 to 244. I don't think we want to orthogonalise when it comes to FIR models, so these lines should be turned off for FIR models, e.g. based on information stored in xBF.name.
Best
Helmut
|