A complete sufficient statistic for the linear model under normality and a singular covariance matrix
Justus Seely · Communication in Statistics- Theory and Methods · 1978
Assuming normality, a general form of a complete sutiicient statistic for the linear mode] having a singular covariance matrix is determined. The relationship between the complete sufficient statistic and some of the methods for determining best linear unbiased estimators given in the literature is also discussed.