# sol10 - \$ m& \$ yx p Q& x p m p& x p ~ | i m u Q um...

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: \$ m & \$ yx p Q & x p m p & x p ~ | i }{ m u { Q um z Y g p itC g p and . Now . The MSE is With this is is . By simulation, The mle has MSE about .015, substantially smaller than the nonparametric plug-in. g d g v v Tv Ty8Q l m 8 5w xw v Q rd ug h m t p ig h m p o nI Pl0 j h g e qSsird )q 8 E p v ki UWf ` \$ d h q \$ x X X & x ` Th y` S v BE 9 f10 \$ Q x b x ` ` v \$ x ` x y` v \$ C PI0 HFDB@ 9 v t ( & r 8 GEC A b3 y ywY 210 Tu)\$ '% s" UTS x ` x ` 3 R p h q ig feSc d b3 aY ` 3 X (1b) is the mean, which, for a is . (1c) The nonparametric plug-in estimator is Q T YBWUTS ` XV X R Q C PI0 HFDB@ 9 ( & 8 " GEC A 3 3 7654 210 )\$ '% #! Homework 10 Solutions where is the smallest data point and hood is maximized by taking If then the observation contributes a factor of it makes the product 0. Hence, The likelihood function is (2) We need the Fisher information matrix for { z U { z And (1) Let . The density is 1 T TFu Y is the largest data point. The likeli- . The mle is otherwise 8 I P v Q p g E 8yQ \$ { z & { Q uh z p n v H p E \$ v )qb3 E p ` 3 EGp w FDB p v A E C@ @ x b 9 8 ` m6 42 0 " Q #! 7531 g ( )p | g d " #! Qm 8yQ \$ & \$ { | " Q m 8Q \$ & g \$ d { \$ & d ' % #! \$ & \$ Q 8Q g & i e D\$ d m { e \$ d z m z p zX ` X & p { z X ` X & p \$ { 8Q & uh z p 3 3 ` Q m m 8 { U { z \$ & 8 m Q 8 \$ 8 { z m { m z p h ` 8 U z p h ` An approximate The estimates standard error is The asymptotic standard error of The inverse Fisher information is Since (3) we get (4) The mle is and and confidence interval is , the maximum data point. Note that is 2 R . The mle is . The gradient of , we have . Solving for where . Hence, z %2 8 q 0p g 1 ( ( 8 q )p g 8 3 p g 8 C g ( ` ` z z H p g p ` p p z \$ ` h ` T} T S Q B )p g T p g \$ d h #\$ p g d h E l o \$ " % x E l o p g x E l o 0 ' &9 EC \$ \$ ! h | Q d ` b0 ` H p so . The ARE is and p g (b) Let dard error of is 95 per cent confidence interval is | | d h ` d TS ` H p g ! H )p \$ g p p #C qT p C 7p ` p h " #T #T 7! #! @ @ " p 3p g TS \$ p )p g \$ !T WW g p 3 8 C ` g p 9 & " U . (5) so R and m Q g P I h U m m Q p m (c) has mean numbers. (d) Note that (e) By the law of large numbers, converges in probability to . So converges in probability to . The true value of is . For an arbitrary distribution , we have so the mle is inconsistent. On the other hand, is still consistent. Thus, d as (6a) . The mle is . Then, . Consistency follows from the weak law of large 3 . . The estimated stan. An approximate p g ...
View Full Document

## This note was uploaded on 05/25/2008 for the course STAT 36625 taught by Professor Larrywasserman during the Fall '01 term at Carnegie Mellon.

Ask a homework question - tutors are online