関西大学学術リポジトリ >
0900 大学院の研究プロジェクト >
私立大学戦略的研究基盤形成支援事業 >
社会的信頼システム創生センター >
社会的信頼学 >
社会的信頼学-第2号 >

Please use this identifier to cite or link to this item: http://hdl.handle.net/10112/8447

Title: 自殺における相対リスクのベイズ推定-経験ベイズ法と階層ベイズ法による縮約推定量の構成-
Other Titles: Application of Bayesian Method to Estimating Relative Risks of Suicide in Japan -Making Shrinkage Estimators of Standard Moratality Rate Using Empirical Bayesian Models and Hierarchical Bayesian Models-
Authors: 紺田, 広明
Author's alias: KONDA, Hiroaki
Keywords: 自殺
標準化死亡比
階層ベイズ法
MCMC
小地域推定
suicide
hierarchical Bayesian models
MCMC
small area estimation
Issue Date: 31-Mar-2014
Publisher: 関西大学社会的信頼システム創生センター
Shimei: 社会的信頼学
Volume: 2
Start page: 1
End page: 34
Abstract: 効果的な自殺対策を行うために,地域における自殺実態の把握の重要性が認識されてきている.地域の自殺実態を表す統計指標として,相対リスクである標準化死亡比がしばしば用いられる.しかし,標準化死亡比は,人口が少ない地域において変動が大きい欠点がある.そのため,信頼性ある推定量が求められてきた.ここでは,最尤推定量,経験ベイズ推定量,階層ベイズ推定量の3種類の相対リスクの推定量を扱うこととした.兵庫県でのベイズ推定では,事前分布にガンマ分布を仮定した相対リスクの基本となるPoisson-Gammaモデルを用い,事前分布を仮定するベイズ推定の枠組みが縮約推定量を構成することを説明した.また,マルコフ連鎖モンテカルロ法による階層ベイズ推定量の構成の仕方を検討し,それぞれの推定量の特長についての整理を行った.Japanese suicide rate got drastically higher in 1998 and it is keeping high since then. And prevention of suicide is getting a most important political issue in contemporary Japan. In terms of effictive preventions of suicide, assessing suicide risks in small areas have been regarded as an essential task, and standardized mortality rates(SMR) is generally used to showing pictures of conditions of suicide in small areas. However, SMR has disadvantages of showing excess variability in calculating it for smaller population areas such as rural districts.Therfore, getting reliable estimators of SMR is one of the most important tasks for suicide researchers and policy makers until now. In this study, we focused on maximum likelihood estimators, empirical Bayesian estimators, and hierarchical Bayesian estimators of SMR as indicators of relative suicide risks. First, we applied basic Poisson-Gamma model for areas in Hyogo prefecture in Japan and discussed that the Poisson likelihood model with Gamma prior distribution constructed the shrinkage estimators. Second, we showed how to construct hierarchical Bayesian estimator with Markov chain Monte Carlo methods. Finally, we provided a comprehensive discussion of these three estimators' features.
type: Departmental Bulletin Paper
URI: http://hdl.handle.net/10112/8447
ISSN: 21868646
Appears in Collections:社会的信頼学-第2号

Files in This Item:

File Description SizeFormat
KU-1100-20140331-01.pdf860.34 kBAdobe PDF
View/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

 

Valid XHTML 1.0! DSpace Software Copyright © 2002-2010  Duraspace - Feedback Terms and Conditions of Use PAGE TOP