1993 年 44 巻 3 号 p. 263-268
In this paper, we examine the small sample properties of R2 based on the Stein-rule estimator of coefficients (say, R2S) when relevant regressors are omitted in a specified model. The following is shown, when the model is correctly specified, the bias of R2S is smaller than that of R2 based on the OLS estimator (say, R2S), and the mean square error (MSE) of R2S is smaller than or comparable with that of R20. But, as the magnitude of specification error increases, both bias and MSE of R2S become larger than those of R20.