TESTING LINEAR RESTRICTIONS ON COEFFICIENTS IN A LINEAR-REGRESSION MODEL WITH PROXY VARIABLES AND SPHERICALLY SYMMETRICAL DISTURBANCES

in this paper we consider the power function of the classical F test f or linear restrictions on the coefficients in a linear regression mode l with spherically symmetric disturbances when proxies are used in the place of unobservable regressors. To illustrate the results we numeri cally evaluate the power function for two families of variance-mixing distributions in a simple univariate regression model. Our results sho w the effects on the power function of the degree of nonnormality and of the correlation between the omitted and the proxy variable.