It would be great if you could add a roll_mad (for median absolute deviation : _median(abs(Xi - median(X)))_ ) in the functions of your package.