Question

Let X1,...X be i.i.d with density f()(1/0)exp(-/0) for r >0 and 0> 0. a. Find the pitman estimator of 0 b. Show that the pitman estimator has smaller risk than the UMVUE of when the loss function is (t-0)2 02 Suppose C. f(x)= 0exp(-0x) and that 0 has a gamma prior with parameters a and p, find the Bayes estimator of 0 d. Find the minimum Bayes risk e. Find the minimax estimator of 0 if one exists. 1

Answer 1

ANSIAER: X,, be r..d with density Let Wolcxp o) . -fon o and 00 (a) (0- -) ohere is porometer to be etirnatedl and les function is (ia-o1) ntc), CERI % 9co) - 0+C , ce R'} 9 9ela) atC, cER' The invamant egtimatos 30x) minimizes Lt)-)f (1-e, Xn)de -d n-e)dG 4(x,-0, xn-o) do n-0)d ezp 02 exp 9d)

(b) Pitman estimatoy hal gmalle niSK than the UMVUE , loss punction is CE-2 X compete and Supficient ond e; xis the UMVUE foo t Seems the Teasonable that 1s UMVUE f0T O X 1,0) X 9 ez ponential ate o and their d z~An,o) tet ne [z] E (t-0)2 n-0Z 2 zP _ m--92z e. n) C m-9-02 na n-1) Dlt-0)2 dz e 0 (t-o mo 2(1) mo -e (t-o)2 UMYUE å 0 is the

(x0-)dxa (Xx)f () f() a omd p 9 ha the gamma prios wi th pavameters The poio density LIkehood function is f 00 LCO) otherwise 9C0) pto) O) = Otheaoise 0 Je'sLodo e f) he posteoio poobabllity olensity tunction Jecco The Bayes estimatoy of 0 iS d) minimum Bayes Risk:- The posteion podbablity of the transla tion E. then P (E/F]*zP(aH (E,F) EleExP(otE,F) be constont o times, length 0 hypothes1s iS E' Than,

G (GE0' &9j (E, e') GJEN 91E' u(E')SE) wEN ewe can eunite as peT gin fonction, cCE wlel) (E) P(EIFJ ECE wEN wcel) E PCEIF) E wEN ECEW bayes &ayes RISK decoder on lattices be The minimum E 0T9 max wEN one is fos minimum eoor and anothed for the decoding Scale factor fo minimum Bayes tuning RISK e Minimum estimatoo Baye estimatur foo o iS not a under Squaved eovoT n0cation mode densit TCO) 1 TC0) do distribution ig posteaios kehood x impeoior poior fouma pcsteior 1 exp END