Behaviormetrika
Online ISSN : 1349-6964
Print ISSN : 0385-7417
ISSN-L : 0385-7417
Articles
AN MCMC APPROACH TO ESTIMATING THE WANDERING IDEAL POINT MODEL FOR “pick any/n” TYPE OF BINARY DATA
Hiroshi Hojo
著者情報
ジャーナル 認証あり

2011 年 38 巻 1 号 p. 97-124

詳細
抄録

The wandering ideal point (WIP) model developed by De Soete, Carroll and DeSarbo in 1986 has been used to analyze individual preferences data. Usually the marginal maximum likelihood (MML) method is applied to estimate the subject's ideal points in the WIP model. However, each ideal point estimated by the Bayes expected a posteriori estimation procedure in the MML method is different in nature from that which should be estimated as a mean vector of the multivariate normal distribution each independently assigned to its respective subject point according to the original WIP model. In this paper I take a Bayesian approach to estimate the distribution of each ideal point. Explicitly, the Gibbs sampler, one of the Marcov Chain Monte Carlo methods, is implemented for fitting the WIP model to “pick any/n” type of binary data. Three applications are provided which demonstrate that the method is useful to estimate the distribution of each subject point.

著者関連情報
© 2011 The Behaviormetric Society
前の記事
feedback
Top