{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

57 this looks just like the variance of a linear

Info iconThis preview shows pages 57–64. Sign up to view the full content.

View Full Document Right Arrow Icon
57
Background image of page 57

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
This looks just like the variance of a linear combination a 1 W 1 a 2 W 2 with a 1 1 , a 2  2 , Var W 1 c 11 , Var W 2 c 22 , and Cov W 1 , W 2 c 12 . 58
Background image of page 58
EXERCISE :Let be a p 1 parameter vector with n ̂ n d Normal 0 , C  Define exp 1 2 ... p ̂ n exp ̂ n 1 ̂ n 2 ̂ np Show that Avar n ̂ n  2 j 1 p h 1 p c jh . 59
Background image of page 59

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
5 . Asymptotic Equivalence and Asymptotic Efficiency We need a way to choose among consistent estimators. Two estimators can be consistent but one can have a distribution more tightly centered about the parameter it is estimating. Unlike finite sample analysis, where we compare variances of unbiased estimators, we have to settle for comparing asymptotic variances. Here we focus only on estimators that are n -consistent. 60
Background image of page 60
DEFINITION : Two sequences of n -consistent estimators for Θ , ̂ n : n 1,2,. .. and ̃ n : n are asymptotically equivalent (more precisely, n - asymptotically equivalent )if n ̃ n ̂ n n ̃ n n ̂ n o p 1 . By the asymptotic equivalence lemma, estimators that are n -asymptotically equivalent have the same asymptotic distribution (when scaled by n ). When n ̂ n and n ̃ n are asymptotically normal – the case we usually consider – asymptotic equivalance means n ̂ n and n ̃ n have the same asymptotic variances. 61
Background image of page 61

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
EXAMPLE : Consider estimating the mean from a population with mean , variance 2 , using random sampling. Take ̂ n X ̄ n and ̃ n a n X ̄ n where a n : n 1,2,. .. is a nonrandom sequence with a n 1. Then n ̃ n X ̄ n n a n 1  X ̄ n Because X ̄ n p , asymptotic equivalence between the two estimators is ensured by n a n 1 o 1 which shows that a n must converge to unity sufficiently quickly. 62
Background image of page 62
In some previous examples, a n 1 c / n for some constant c , and so n a n 1 c / n 0. But, say, a n 1 cn 3/4 would work, too. There are uncountably many other choices. These kinds of examples again illustrate how one can abuse asymptotic analysis. For a given sample size n , we could choose a n very far from one – by choosing c large in magnitude – and just assert that, as n , a n will approach one. If possible – and it rarely is in complicated situations – we can use unbiasedness, or look at mean squared error, to limit abuse of asymptotics.
Background image of page 63

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Image of page 64
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

Page57 / 77

57 This looks just like the variance of a linear...

This preview shows document pages 57 - 64. Sign up to view the full document.

View Full Document Right Arrow Icon bookmark
Ask a homework question - tutors are online