On the use of some extraneous information in the estimation of the coefficients of regression