group analysis - how to estimate smoothness when there is no noise term

Hi everyone,

I understand that 3dClustSim is based on the spatial smoothness measured from the GLM residual term. In case one wants to run group analysis on some other kind of data, e.g. SVM weight maps, where the residual term does not exist, what is the suggested way to obtain an estimate of smoothness? Is it valid to apply the same smoothness (calculated from GLM)?

Thanks a lot.

Both the smoothness and the correction are some sort of approximations. If you can manage to provide some rough estimate for spatial relatedness, it might be good enough.