PUKYONG

이중 다단계 일반화 선형모형을 이용한 강건한 연령-기간-코호트 분석

Metadata Downloads
Abstract
Data analysis studies on deaths due to disease have considered the trend of time in the past. However, in recent years, more studies are being conducted considering changes in cohort as distinct from age and period.
In this paper, based on the usefulness of the cohort effect, we attempted to examine the expansion of the age-period-cohort (APC) analysis model to double hierarchical generalized linear model (DHGLM) for an effective and robust fit. Through this, the advantages of APC analysis using DHGLM and the strengths of the model were reviewed, and the usability was improved by developing related R codes. As a result, the advantages and expected effects of a regression model with random effect in APC analysis are examined.
Author(s)
고영규
Issued Date
2021
Awarded Date
2021. 2
Type
Dissertation
Keyword
Age-period-cohort analysis Generalized linear mixed models framework Double hierarchical generalized linear model
Publisher
부경대학교
URI
https://repository.pknu.ac.kr:8443/handle/2021.oak/2357
http://pknu.dcollection.net/common/orgView/200000365534
Affiliation
부경대학교 대학원
Department
대학원 통계학과
Advisor
노맹석
Table Of Contents
Ⅰ. 서론 1
Ⅱ. APC 분석 7
2.1 APC 분석의 필요성 7
2.2 모형 리뷰 11
2.2.1 Reduced two-factor models 13
2.2.2 Constrained generalized linear models 16
2.2.3 Nonlinear parametric transformation 18
2.2.4 Proxy variables 19
2.2.5 Generalized linear mixed models framework 20
Ⅲ. 연구 모형 25
3.1 연구 모형 소개 25
3.2 GLM 27
3.3 Joint GLM 28
3.4 HGLM1 29
3.5 HGLM2 32
3.6 DHGLM 32
3.7 다단계 가능도 33
3.8 모형 평가 36
3.9 과산포 모형을 위한 이중지수족 39
Ⅳ. 연구 자료 및 분석 40
4.1 연구 자료 개요 40
4.2 분석 제약 45
4.3 모형 적합 46
4.3.1 GLM 48
4.3.2 Joint GLM 52
4.3.3 HGLM1 56
4.3.4 HGLM2 60
4.3.5 DHGLM 64
4.4 잔차 도표 68
4.5 적합 결과 요약 72
4.6 시뮬레이션 77
4.6.1 시뮬레이션1 79
4.6.2 시뮬레이션2 83
4.6.3 시뮬레이션3 87
4.6.4 시뮬레이션4 91
4.6.5 시뮬레이션5 95
4.6.6 시뮬레이션6 99
4.6.7 시뮬레이션 결과 요약 103
Ⅴ. 결론 및 제언 106
5.1 결론 106
5.2 시사점 108
5.3 한계점 109
참고문헌 111
부록1 암 사망 분석 자료 118
부록2 모형 적합 R 코드 119
Degree
Doctor
Appears in Collections:
대학원 > 통계학과
Authorize & License
  • Authorize공개
Files in This Item:

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