Asymptotic Distribution of Deviance Statistic in Logistic Regression

• 발행기관: 한국자료분석학회
• 저자명: Jung Hee Jeong Kwang Mo Jeong
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.5 No.4, 749~759쪽, 전체 11쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2003.12.30

영문 초록

The deviance statistic is used as a goodness-of-fit measure for comparing the current model against the saturated model. Saturated models are defined differently even for the same data according to the methods of data input. The usual statistical packages provide deviance statistics in different ways according to the assumed saturated models. Two types of data input formats will be explained. Through a Monte Carlo study we also check the asymptotic distribution of deviance statistics. Under certain regularity conditions the deviance statistic is asymptotically χ² distributed, and hence this can be useful for testing the adequacy of the fitted model. Unfortunately for continuous predictors the deviance statistic is not guaranteed to be χ² distributed.

목차

1. Introduction

2. Review on Logistic Regression Model

3. Deviance Statistic

4. Monte Carlo Study

5. Summary

References

키워드

logistic regression model saturated model deviance asymptotic distribution