Thanks for your response,
I found where the problem was... In fact when SPM receive a matrix in the "multiples regressors" entry, it does consider all regressors in it as covariates and automatically detrend all of them. Inside the spm_detrend function, the polynomial adjustment was the problem. When skipping this, I now have the same results with both models.
However, I'm now wondering how SPM deals with the variance associated with the mean signal intensity that would otherwise be assigned to an experimental condition. Does it remove the mean value at a further step?
Elsa
|