What accuracy statistics really measure

Abstract

Provides the software estimation research community with a better understanding of the meaning of, and relationship between, two statistics that are often used to assess the accuracy of predictive models: the mean magnitude relative error (MMRE) and the number of predictions within 25% of the actual, pred(25). It is demonstrated that MMRE and pred(25) are, respectively, measures of the spread and the kurtosis of the variable z, where z=estimate/actual. Thus, z is considered to be a measure of accuracy, and statistics such as MMRE and pred(25) to be measures of properties of the distribution of z. It is suggested that measures of the central location and skewness of z, as well as measures of spread and kurtosis, are necessary. Furthermore, since the distribution of z is non-normal, non-parametric measures of these properties may be needed. For this reason, box-plots of z are useful alternatives to simple summary metrics. It is also noted that the simple residuals are better behaved than the z variable, and could also be used as the basis for comparing prediction systems