The Hot Air in R² and Consistent Measures of Explained Variation

Abstract

Abstract The size of R 2 and R̅ 2 are poor specification indicators since correctly specified models can have “low” R 2 values and misspecified models often have “high” R 2 values. Further, R 2 and R̅ 2 , as typically calculated, are found to be inconsistent measures of goodness of fit whenever the mean of the dependent variable is not stationary. Consistent estimation methods are proposed.