I'm not sure what kind of absolute values you are thinking of. The values provided are distances from a reference surface. The surface we use as the reference is the mean surface of the cohort which is an unbiased surface and hence has the zero mean property you are noticing. I don't think there is an unbiased surface that would give "absolute" values.
To investigate correlations of behaviour and shape I would advise you to create a design matrix containing the demeaned behavioural values and then perform vertex analysis using this design matrix. This will then show you values at each surface location.
All the best,
Jodi Gilman <[log in to unmask]> wrote:
I performed a vertex analysis comparing two groups, and found some interesting differences. I wanted to correlate scalar values form individual subjects with behavior. I extracted the data by choosing a max voxel from the output of randomise, and, using fslmeants, pulled the data from each subject. However, when looking at the data, it appears that the individual values are relative to the other group, and not absolute (for example, the mean of Group 1 is the negative value of the mean of Group 2).
Is there a way to get an absolute scalar value from each subject at a particular location?