ABSTRACT
This chapter discusses the problem of estimation in a general linear model with purely random coefficients when the coefficients are contemporaneously correlated; and autocorrelated. It assumes that the ki uncorrelated with xki. The operational estimate of can be obtained by using either the estimated variance-covariance matrix Do obtained by using the least squares estimates of kks or the estimated variance-covariance matrix Do obtained by using the operational generalised least squares estimates of kks in place of the true matrix Do in the generalised least squares estimator. The chapter note that d2 consists of nK2 unknown elements which are interrelated with each other. An operational estimator of can be obtained if people use an estimated variance-covariance matrix in place of the unknown matrix of the generalised least squares. The important point to note is that all other operational estimators of o, obtained by using an estimated dispersion matrix whose elements are obtained, will have similar large sample properties.
