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

Bayesian multivariate binary respons...

Hahn, Eugene David.

Linked to FindBook

Google Book

Amazon

博客來

Bayesian multivariate binary response models and their application to multiple response data.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Bayesian multivariate binary response models and their application to multiple response data./
Author:	Hahn, Eugene David.
Description:	222 p.
Notes:	Source: Dissertation Abstracts International, Volume: 62-02, Section: A, page: 0678.
Contained By:	Dissertation Abstracts International62-02A.
Subject:	Statistics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3004031
ISBN:	0493130128

Bayesian multivariate binary response models and their application to multiple response data.
Hahn, Eugene David.

Bayesian multivariate binary response models and their application to multiple response data. - 222 p.

Source: Dissertation Abstracts International, Volume: 62-02, Section: A, page: 0678.

Thesis (Ph.D.)--The George Washington University, 2001.

Advances in Markov chain Monte Carlo techniques have made the estimation of complex high-dimensional models more tractable, especially in the Bayesian approach to inference. One class of models that has been the subject of recent attention is the class of multivariate binary response models, of which the multivariate probit and the multivariate logit models are frequently encountered special cases. An important decision that arises when using these models is the selection of the link function. In previous work, Chambers and Cox (1967) identified situations in which the fitting of univariate binary response models was substantively impacted by the differential selection of the probit or the logit link function. However, little is known about how the choice of link function affects model fit in multivariate binary response models.

ISBN: 0493130128Subjects--Topical Terms:

517247
Statistics.

Bayesian multivariate binary response models and their application to multiple response data.
LDR:02978nmm 2200325 4500 001 1858279
005 20040927073704.5
008 130614s2001 eng d
020 $a 0493130128
035 $a (UnM)AAI3004031
035 $a AAI3004031
040 $a UnM $c UnM
100 1 $a Hahn, Eugene David. $3 1945975
245 1 0 $a Bayesian multivariate binary response models and their application to multiple response data.
300 $a 222 p.
500 $a Source: Dissertation Abstracts International, Volume: 62-02, Section: A, page: 0678.
500 $a Director: Refik Soyer.
502 $a Thesis (Ph.D.)--The George Washington University, 2001.
520 $a Advances in Markov chain Monte Carlo techniques have made the estimation of complex high-dimensional models more tractable, especially in the Bayesian approach to inference. One class of models that has been the subject of recent attention is the class of multivariate binary response models, of which the multivariate probit and the multivariate logit models are frequently encountered special cases. An important decision that arises when using these models is the selection of the link function. In previous work, Chambers and Cox (1967) identified situations in which the fitting of univariate binary response models was substantively impacted by the differential selection of the probit or the logit link function. However, little is known about how the choice of link function affects model fit in multivariate binary response models.
520 $a New multivariate binary response models were proposed, and a Bayesian approach to the estimation of these models was adopted. Differences in model fit were predicted to be influenced by three factors: the presence of extreme independent variable levels, high dependent variable correlation, and sample size. Model fit was assessed using two methods of computing Bayes factors as well as the Deviance Information Criterion (DIC). A Monte Carlo study was conducted to examine the effects of these three factors.
520 $a The presence of both extreme independent variable levels and high dependent variable correlation were found to affect the difference in fit afforded by the two link functions. Additionally, it was found that marginal models and random effects models have different “preferred link functions”, for which model fit is more generally enhanced. The results with respect to the sample sizes that were considered in the current program of research were more equivocal.
520 $a As a follow-up study, the models were then estimated using a real-world marketing research data set. The findings from this analysis corroborated those of the Monte Carlo study.
590 $a School code: 0075.
650 4 $a Statistics. $3 517247
650 4 $a Business Administration, Marketing. $3 1017573
650 4 $a Sociology, Theory and Methods. $3 626625
690 $a 0463
690 $a 0338
690 $a 0344
710 2 0 $a The George Washington University. $3 1017405
773 0 $t Dissertation Abstracts International $g 62-02A.
790 1 0 $a Soyer, Refik, $e advisor
790 $a 0075
791 $a Ph.D.
792 $a 2001
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3004031