Dear Hongyu,
> in our fMRI study, the subjects need to press button to indicate
> their feeling
> during each trial. I think we can do a basic correlation analysis
> with the
> behavioral result. My question is, after finishing basic analysis,
> we are trying
> to do a group correlation analysis. Do we need to enter the average
> feeling for
> each subject as an extra EV which will be orthogonalised wrt the
> group mean
> EV? I am not sure about this group correlation analysis.
as part of the output from each subject you will have a "cope", i.e. a
"contrast of parameter estimates". In your case this will be a map of
regression coefficients determining the "correlation" with the
"feeling regressor". These will be the inputs for a 2nd level
analysis. If you set up your 2nd level design as a one sample t-test
you will for each voxel get the average (across the subjects)
correlation with the "feeling regressor". Hence, these is no need to
enter a "feeling regressor".
Good luck Jesper
|