Always resample the original number of observations