A STUDY OF VARIOUS IMPUTATION METHODS

Abstract

In most samples, even with careful design, values are commonly missed. When the existing data is missing too many values, the analysis can be incorrect. Suitable treatment is needed to avoid such cases. Two principal approaches to estimate with missing data are weighting and imputation. Weighting typically is used in unit non-response which is viewed as the inverse of the response probabilities associated with the response mechanism. Imputation is used in item non-response. The imputed values are sample-based. So imputation is the solution widely used to prohibit them. There are a variety of imputation methods. The goal of any imputation technique is to produce a complete data set, which can be analyzed using complete-data inferential method. In this paper, we propose several methods of imputation, and compare their efficiencies through the case study.;우리가 실제로 분석하기 위해 얻는 대부분의 data set은 결측치를 많이 포함하고 있으며, 이로 인해 기존의 분석 방법이나 추론 방법을 통해 얻는 결과가 부정확해질 수 있다. 완벽한 data set을 얻기 위해서 적절한 조치가 필요한데, 가장 많이 쓰이는 방법으로 weighting과 imputation이 있다. imputation에는 여러 방법론이 있는데, 이 논문에서는 크게 deterministic방법과 stochastic방법을 실제 데이터 분석을 통해 비교해 보았다.