Working Papers, University of California, Davis, Department of Economics 00-7
The distribution of the least squares and the instrumental variable estimators of the coefficients in a linear relation is noncentral student when the data are normally distributed around possibly non-constant means. This is the claim of the paper and we show for what definition of a noncentral student density this claim to compact summary of the vast literature is justified. Unfortunately, the definition of the noncentral student density is as complicated as its parent, the noncentral Wishart density. Both are defined in terms of infinite series of zonal polynomials of all orders. We have developed however a recursive online method that generates these polynomials sequentially ad infinitum for bivariate and trivariate densities. The time is here that the practicality of the theory can be widened considerably.