東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Estimation of high dimensional predi...

Xu, Xinyi.

Linked to FindBook

Google Book

Amazon

博客來

Estimation of high dimensional predictive densities.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Estimation of high dimensional predictive densities./
Author:	Xu, Xinyi.
Description:	65 p.
Notes:	Source: Dissertation Abstracts International, Volume: 66-06, Section: B, page: 3216.
Contained By:	Dissertation Abstracts International66-06B.
Subject:	Statistics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3179839
ISBN:	0542201119

Estimation of high dimensional predictive densities.
Xu, Xinyi.

Estimation of high dimensional predictive densities. - 65 p.

Source: Dissertation Abstracts International, Volume: 66-06, Section: B, page: 3216.

Thesis (Ph.D.)--University of Pennsylvania, 2005.

Let X | mu &sim; Np(mu, vxI) and Y | mu &sim; Np(mu, vyI) be independent p-dimensional multivariate normal vectors with common unknown mean mu. Based on only observing X = x, we consider the problem of obtaining a predictive density p(y | x) for Y that is close to p( y | mu) as measured by expected Kullback-Leibler loss. We connect this problem with estimating a multivariate normal mean under quadratic loss by establishing a new link identity, which explains striking similarities between these two problems and can be used to obtain analogous unifying theories of minimaxity and admissibility for density prediction. The natural straw man for this problem is the Bayesian predictive density pU( y | x) under the uniform prior pi U(mu) &equiv; 1, which is best invariant, minimax and admissible when p = 1 or 2. We show that any Bayes predictive density will be minimax if it is obtained by a prior yielding a marginal density that is superharmonic or whose square root is superharmonic. This yields wide classes of minimax procedures that dominate pU( y | x) including Bayes predictive densities under superharmonic priors. We also prove that all admissible procedures for this problem have to be formal Bayes rules, and that the conditions in Brown and Hwang (1982) are sufficient for a formal Bayes rule to be admissible. Furthermore, we show that many of these results can be extended to the linear regression setting with slight modifications. The shrinkage and multiple shrinkage properties of Bayesian predictive densities are also discussed.

ISBN: 0542201119Subjects--Topical Terms:

517247
Statistics.

Estimation of high dimensional predictive densities.
LDR:02404nmm 2200265 4500 001 1819551
005 20061005085933.5
008 130610s2005 eng d
020 $a 0542201119
035 $a (UnM)AAI3179839
035 $a AAI3179839
040 $a UnM $c UnM
100 1 $a Xu, Xinyi. $3 1908828
245 1 0 $a Estimation of high dimensional predictive densities.
300 $a 65 p.
500 $a Source: Dissertation Abstracts International, Volume: 66-06, Section: B, page: 3216.
500 $a Supervisor: Edward I. George.
502 $a Thesis (Ph.D.)--University of Pennsylvania, 2005.
520 $a Let X | mu &sim; Np(mu, vxI) and Y | mu &sim; Np(mu, vyI) be independent p-dimensional multivariate normal vectors with common unknown mean mu. Based on only observing X = x, we consider the problem of obtaining a predictive density p(y | x) for Y that is close to p( y | mu) as measured by expected Kullback-Leibler loss. We connect this problem with estimating a multivariate normal mean under quadratic loss by establishing a new link identity, which explains striking similarities between these two problems and can be used to obtain analogous unifying theories of minimaxity and admissibility for density prediction. The natural straw man for this problem is the Bayesian predictive density pU( y | x) under the uniform prior pi U(mu) &equiv; 1, which is best invariant, minimax and admissible when p = 1 or 2. We show that any Bayes predictive density will be minimax if it is obtained by a prior yielding a marginal density that is superharmonic or whose square root is superharmonic. This yields wide classes of minimax procedures that dominate pU( y | x) including Bayes predictive densities under superharmonic priors. We also prove that all admissible procedures for this problem have to be formal Bayes rules, and that the conditions in Brown and Hwang (1982) are sufficient for a formal Bayes rule to be admissible. Furthermore, we show that many of these results can be extended to the linear regression setting with slight modifications. The shrinkage and multiple shrinkage properties of Bayesian predictive densities are also discussed.
590 $a School code: 0175.
650 4 $a Statistics. $3 517247
690 $a 0463
710 2 0 $a University of Pennsylvania. $3 1017401
773 0 $t Dissertation Abstracts International $g 66-06B.
790 1 0 $a George, Edward I., $e advisor
790 $a 0175
791 $a Ph.D.
792 $a 2005
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3179839