The spatial regression is very similar to averaging the timecourses of the voxels within a spatial map (using the values >within the map as weightings); the big difference though is that this is done as a multiple regression

(using all ICA spatial >maps as regressios in one big GLM), with all that that implies,

rather than one at a time separately.

Stephen thanks for your response.

So, we have 36 components for our temporal concatenated group ICA. With the spatial regression, our multiple regression will consist of 36 regressors? The values of the regressors are Z-scores? IF the Z-scores are the "x-values/independent variables", what exactly is the "y-value/dependent variable" Is this regressed on a voxel-by-voxel basis where the final result is an estimated 36 time-courses/components per subject?

Thanks,

Michael