Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site noscvax.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!whuxl!whuxlm!harpo!decvax!ittvax!dcdwest!sdcsvax!noscvax!wjohnson
From: wjohnson@noscvax.UUCP (Roger W. Johnson)
Newsgroups: net.math.stat
Subject: Admissibility & Exponential Families
Message-ID: <897@noscvax.UUCP>
Date: Thu, 25-Apr-85 14:12:48 EST
Article-I.D.: noscvax.897
Posted: Thu Apr 25 14:12:48 1985
Date-Received: Sat, 27-Apr-85 06:17:00 EST
Reply-To: wjohnson@cod.UUCP (Roger W. Johnson)
Organization: Naval Ocean Systems Center, San Diego
Lines: 29





Let the variate X have exponential density

  f(x/w) = c(w) s(x) exp( m(w)t(x) )          a < x < b

and suppose that the estimate d(x) of m(w) incurs squared error loss

  L(d,m) = (d(x) - m(w))^2.

Can someone supply me with a reference which summarizes work done in
answering the question "When is d(X)=X an admissible estimate of m(w)?"

I am especially interested in the special case where X has a density
which belongs the four parameter Pearson class where only the mode
is unknown.  Thanks in advance for any information you can send along.


Roger Johnson/Naval Ocean Systems Center/Code 632/San Diego, Ca 92152




-- 
Roger Johnson
2150 Pacific Beach Drive, Apt 207
San Diego, Ca     92109