 |
 |
Dear SPM users, For my event related fMRI analysis, I use a set of 13 FIR basis functions (with blocks of 2.5 seconds in order to be able to calculate the neuronal response for the duration of the whole theoretical HRF for every voxel). To be able to plot the right neuronal reponses for each voxel, I thought it was necessary to omit the orthogonalization part in the fMRI analysis (spm_fmri_design.m). If I do that, the neuronal response can simply be plotted by all the beta's estimated in the model since the basis functions are all 1 (usually for the neuronal response the following formula hold: response=Sum(from1:n) (beta_n*basisfunction_n, in which n=total number of basis functions). However, by skipping the orthogonalization part, the different columns in the design matrix that form my predictors are still correlated. I wonder how this affects the betas estimated from my model. I would really like to know how this influences my results?! Can anybody please give me some answers or advice? Thank you in advance! Regards, Debby