조사 자료의 품질 검증 연구: 측정오차를 중심으로

Abstract

조사 자료는 보통 오차를 포함하고 있으며 오차의 정도가 심각한 자료를 사용하여 통계분석을 한다면 심각하게 편향된 분석 결과를 도출하게 된다.

이 연구는 표본조사 자료에서 발생할 수 있는 오차를 살펴보고, 그중에서 측정오차에 대해 심도 있게 살펴보는 것에 목적을 두고 있다. 표본조사 자료에 측정오차가 있는 경우에는 보정하는 방안도 다양하게 모색하여 제안하였다. 또한, 주요 해외 표본조사 자료에서의 측정오차 관리 방안 사례를 조사하여 고품질 자료 생산을 위한 조사 자료 관리 방안을 마련하였다.

Survey data are usually constructed through sampling by extracting some subjects from the population of interest. Since we make inferences about the entire population with the survey data, there will be a difference between the sample estimate and the true population value. The difference between the sample estimate and the true population value is defined as an error, and the error can occur by various causes and situations. The purpose of this study is to examine the quality of survey data and to suggest ways for quality improvement. This study analyzes the measurement errors occurred in the sample survey data, and proposes various methods for correcting them. Main results of this study are as follows: Measurement error correction methods were performed by using the survey data. Measurement error correction methods include measurement error correction in the linear regression models with continuous outcomes, kernel density estimation for heaped data, and fractional hot deck imputation method. An R package that implements measurement error correction methods for regression models with continuous outcomes is the function ‘mecor’ in the package ‘mecor’. An R package that implements measurement error correction methods for kernel density estimation for heaped data is the function ‘dheaping’ in the package ‘Kernelheaping’. An R package that implements measurement error correction methods for fractional hot deck imputation method is the function ‘FHDI’ in the package ‘FHDI’. As an R package has some packages for measurement error corrections, most researchers can use it without difficulty. Meanwhile, a survey data management for producing high-quality data was proposed through a case study of measurement error management methods in major overseas sample survey data. It is fundamentally difficult to eliminate measurement errors. However, this can be solved by active participation of the respondents, sincere conduct and appropriate management of the interviewers, and planning of the surveys reflecting the reality.