Hopefully anyone can help me with that:
Is it possible that in spm5 the output from my statistics calculation (t values),
differ significantly depending on whether:
1. I compute my event-related design statistics with the "names durations onsets" in the 'regular' position spm expects them to be:
jobs{1}.stats{1}.fmri_spec.sess(1).multi= 'conditions file' -
and my regressors (orthagonalized to the conditions) at the postion where spm requires the regressors: jobs{1}.stats{1}.fmri_spec.sess(1).multi_reg= 'regressor file'
OR
2. I leave jobs{1}.stats{1}.fmri_spec.sess(1).multi=' ' empty, and add my conditions to the "regressor-position" (jobs{1}.stats{1}.fmri_spec.sess(1).multi-reg), in the following way:
first columns (eg. 6 as six conditions are present) will be my HRF convolved conditions, followed by another six columns of headmotion regressors (orthagonalized to the first six columns (=hrf convolved conditions)
--> LIke this I can maximize the task related effects compared to my other regressors.
FACT is that with version 2 I get much higher t-values (in the range of a t value of 5 becomes 7 e.g. at the same local maximum).
Should they not be both giving equal results?
(the hrf s in both are the same..., for both I use orthagonalized regressors to the hrf convolved conditions, whereas in point 1. I enter them as 'seconds' and convolve them separately, without entering them to the model)
Why does Spm 5 perfrom differently depending on wether I add the conditions (hrf-convolved) together with the regressors (2.) compared to when do add them non convolved the way mentioned in (1.)?
Please can someone help me, I dont know which output is the correct one- as to me both should give the same results-
Natalie
|