Using empirical partially Bayes inference for increased efficiency (Q1081237)

The author shows how empirical partially Bayes methods can be utilized as a means for improving the efficiency of an estimator of the structural parameter in a class of problems in which the number of nuisance parameters increases to infinity. It is proposed to estimate the structural parameter in an asymptotically unbiased way and apply James- Stein shrinkage to the nuisance parameter estimators. It is shown that when the shrinkage estimators are carefully chosen, this yields estimators generally more efficient than the maximum likelihood one.