Is dominating Positive Part James Stein estimator when estimating the mean of a multivariate normal of dimension 3 with known variance(all equal) an open problem?

If not, what is this estimator called? Can you point out some references?


Yes, there is a published reply to this challenge, by Peter Yi-Shi Shao and William E. Strawderman in 1994: https://projecteuclid.org/euclid.aos/1176325640

enter image description here

They provide conditions on estimators of the type

enter image description here

to ensure domination. Interestingly such estimators are also inadmissible and hence are dominated by other (generalised Bayes) estimators (see p. 1530).

  • $\begingroup$ Do we know the admissible Generalized Bayes estimators that dominate JS, or do we just rely on some existence theorem? $\endgroup$ – Cagdas Ozgenc Oct 31 '18 at 10:37
  • $\begingroup$ No, as mentioned on p.1530, there exist generalised Bayes estimators dominating this one, they may belong to the class discussed on p. 1530, if not to (3.1) [which does not contain analytic functions], but no explicit improvement is known. $\endgroup$ – Xi'an Oct 31 '18 at 16:55
  • 1
    $\begingroup$ By the sufficiency principle, one should only consider the sample mean. $\endgroup$ – Xi'an Nov 1 '18 at 11:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.