hi marco,
my understanding is that you simply enter the behavioral rating as an additional predictor in your glm. e.g., when your subjects viewed images of different objects and their task was to make, say, pleasantness ratings, you use one predictor for the image presentation and then add another one that represents the pleasantness ratings. (both of them convolved with the hrf of course). the beta weight for that additional regressor tells you how strongly activation in a given voxel covaries with the behavioral ratings. so there is no need to dichotomize the data as you suggested in your email.
hope this helps!
michael
ps: just yesterday i read an article that seemed to do just that. maybe you want to have a look at it: zahn, moll, krueger, huey, garrido, & grafman (2007).
Am 30.07.2009 um 23:31 schrieb Hubert,Marco:
Dear all,
i know the vul et al. paper but nevertheless i want to know how i can calculate correlations between fmri and behavioral data.
i.e. i have a contrast high versus low and a corresponding vector of evaluations regarding the images within high and low. How can i get a correlation between those data? (more focused: how can i get a correlation between an activated brain regions for high versus low images and the higher evaluations on a specific scale regarding the high images)
Thank you very much!!
Best
Marco
