Dear SPM List,
I have a question concerning why the result of the parametric modulation is different in two (very similar/identical?) ways.
Experiment:
Subjects receive Stimulus A and B. (e.g. two different odors - duration > 6s). Subsequently they have to rate them on a scale. This operation is repeated oftenly (also with different kinds of manipulation).
Now, I want to parametrically model my ratings onto the Stimulus A and B to see if there are regions correlating with the ratings.
As far as i understand I have two options in designing the matrix.
1. I model Stimulus onsets and durations from stimulus A and B separately in two differnt columns each combined with the corresponding rating (paramter) -> I get a column for Stimulus A*rating and a column for Stimulus B*rating.
Contrast: (1 1) -> T-test vs 0
2. I model Stimulus onsets and durations from stimulus A and B together in one column - again combined with the corresponding rating (parameter). I get one column Stimulus*rating
Contrast (1) -> T-test vs 0
I thought the results from both models should be almost identical but they are not. In fact model 1 seems to fit better for some reason. Could it be that putting them alltogether in one matrix inceases the variance to strong? I mean there is a strong difference between Stimulus A and B(effective as well as rated). Or am I completly wrong and something else might be affecting the data.
Any help is appreciated!
Nico
|