Estimation in a Gaussian linear regression model under type II censoring

In the setting of a Gaussian linear regression model, estimation of the regression parameters is considered when a fixed number of design points are replicated in the experiment, and the response data at each design point are type II censored. A judicious approximation of the likelihood equation leads to closed form solutions which are analogous to the maximum likelihood estimators in the uncensored case. The exact and asymptotic properties of these modified maximum likelihood estimators are investigated, and efficiency comparisons are made with the maximum likelihood estimators.