Estimation of linear models with missing data: The role of stochastic linear constraints

Assuming the nonavailability of some observations and the availability of some stochastic linear constraints connecting the coefficients in a linear regression, the technique of mixed regression estimation is considered and a set of five unbiased estimators for the vector of coefficeints is presented. They are compared with respect to the criterion of variance covariance matrix and conditions are obtained for the superiority of one estimator over the other.