능형회귀(Ridge Regression)와 LASSO를 이용한 모델 보조 보정(Model-assisted Calibration) 추정량 연구 = A Study on Model-assisted Calibration Estimator Using Ridge and LASSO Regression

다국어 초록 (Multilingual Abstract)

This study aims to compare the traditional calibration estimator with the model-assisted calibration estimator using ridge regression and LASSO regression. As the response rate in the probability sample decreases rapidly, surveys using non-probability samples that can easily collect data are increasing. However, since there is a risk of bias when using a non-probability sample, it is necessary to select necessary auxiliary variables and use them to calibrate them. This study selected a variable that is highly correlated with the dependent variable through ridge regression and LASSO regression, included it in the model, obtained the estimator through the calibration process, and analyzed which method is effective by considering the bias and mean square error of each of them. For this, the Community Health Survey data of the Korea Disease Control and Prevention Agency were used. As a result of the analysis, the estimation efficiency of the model-assisted calibration method using ridge regression and LASSO regression was better than the traditional calibration method, and the estimation efficiency of ridge regression and LASSO regression was similar.

목차 (Table of Contents)

• 목 차
• 제1장 서 론 ····························· 1
• 제2장 보정 과정 (Calibration Procedure) ············· 4
• 제1절 전통적인 가중치 보정 과정 ················· 4
• 제2절 모델 보조 가중치 보정 과정 ················ 6
• 제3장 능형 회귀 추정 ························ 8
• 제1절 능형 회귀 추정 ······················· 8
• 제2절 능형 회귀 보정 가중치 ··················· 10
• 제4장 LASSO 회귀 추정 ······················ 11
• 제1절 LASSO 회귀 추정 ····················· 11
• 제2절 LASSO 회귀 보정 가중치·················· 12
• 제5장 시뮬레이션 ·························· 13
• 제1절 자료 소개 ·························· 13
• 제2절 시뮬레이션 방법 ······················ 15
• 제3절 시뮬레이션 결과 ······················ 18
• 제6장 결 론 ····························· 23
• 참 고 문 헌 ····························· 25
• ABSTRACT ····························· 27
• <표 목 차>
• <표 5-1> 시뮬레이션에 사용한 모델 ················ 16
• <표 5-2> 모델을 통한 가중치 보정 추정량에 대한 통계량 (전통적인 보정) ································· 18
• <표 5-3> 모델을 통한 가중치 보정 추정량에 대한 통계량 (능형 회귀)··································· 19
• <표 5-4> 모델을 통한 가중치 보정 추정량에 대한 통계량 (LASSO) 19
• ＜표 5-5＞ 지역사회건강조사 가중치 보정 추정량에 대한 통계량 ······· 21