The standard deviation of the error term in the fit of a model to the calibration data. This is a poor measure for comparing the predictive validity of time-series models; use it only if no other measures can be obtained and use it with skepticism.