Hi Roger,

A fixed-cost threshold simply keeps the highest/strongest N-percentile among the ROI-to-ROI connections, so if you want to threshold a correlation matrix Z by a fixed-cost threshold value v, you may for example use something like:

 mask = find(~eye(size(Z,1)));
 N = numel(mask);
 [~,in] = sort(Z(mask));
 in(1:N-round(N*v/2)*2) = [];
 Z_thresholded = nan(size(Z));
 Z_thresholded(mask(in)) = Z(mask(in));

(e.g. for v=.15 threhsold value, the Z_thresholded matrix will contain the Z values for the highest 15% connections, and NaN's for all the rest)

Let me know if this is what you were referring to
Best
Alfonso

Originally posted by Roger Beaty: