The normalized mean square error (NMSE) is taken as the ratio of the mean approximation square error to the variance of the data in question. As an example, a NMSE of 0.1 implies that on the average the approximation error is ten times smaller than the variance of the data.