SMALL SAMPLE PROPERTIES OF R² BASED ON THE STEIN-RULE ESTIMATOR IN A MISSPECIFIED LINEAR REGRESSION MODEL

抄録

In this paper, we examine the small sample properties of R² based on the Stein-rule estimator of coefficients (say, R²_S) when relevant regressors are omitted in a specified model. The following is shown, when the model is correctly specified, the bias of R²_S is smaller than that of R² based on the OLS estimator (say, R²_S), and the mean square error (MSE) of R²_S is smaller than or comparable with that of R²₀. But, as the magnitude of specification error increases, both bias and MSE of R²_S become larger than those of R²₀.