1994 年 45 巻 4 号 p. 306-320
In this paper, following Tiao, Tan and Chang (1977), we consider a constrained bivariate linear regression model and introduce two types of estimators after a preliminary test of independence of two equations. One of them is a pre-test estimator using an unrestricted estimator of covariance matrix and the other is one using a restricted estimator of covariance matrix. We derive their risks and compare sampling performances of several estimators.