A Bayesian Detection Method for Influential Effects

• 발행기관: 한국자료분석학회
• 저자명: Sangjeen Lee Younshik Chung
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.1 No.2, 51~57쪽, 전체 7쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 1999.12.30

영문 초록

In a random effect model, the influential effect means that it is distinct from others and characteristic from other effects. In this paper, we propose a Bayesian method detecting this influential effect with the mean-shift model. Like other Bayesian approach, our proposed method also has the unknown constant problem in Bayes factor from improper prior and the computational confusion. To overcome them, we use Savage-Dickey fractional Bayes factor (SDFBF; Chung and Lee; 1998). Finally, a simulation is performed with a generated data set from the model including an influential effect. And we also analyze a real data set.

목차

1. Introduction

2. Model Specification and SDFBF for Detecting Influential Effect

3. Illustrative Examples

Reference

키워드

SDFBF SDIBF mean-shift model influential effect minimal training sample random effect model