범죄학 연구와 결측 : 실제 자료를 활용한 결측 대치법 비교 분석

Researchers often find that their datasets have missing values. In self-report surveys that are frequently utilized in social science research, respondents are able not to answer certain questions, or accidentally, they may not answer an item. These missing values are relevant to occurrence of type 1 error and type 2 error. The current study reviewed the significance of proper treatment of missing values and discussed several methods to handle missingness in data. In addition, this study compared results from several imputation methods, utilizing data from the Korean Children & Youth Panel Survey. It is found that results from multiple imputation and regression-based imputation are similar to those from the proxy data of real world, whereas the row data that include missing values have some symptoms of type 1 and type 2 errors. Given the results, it is imperative to properly handle missing values in criminology research. In survey datasets, missing values may be related to crime and delinquency so that non-treatment of missing values may result in significant subjects in the dataset (e.g., individuals with high criminality), biasing the estimations of statistical analyses. The results also have some practical implications in that empirical findings are often utilized as scientific evidence of policies and practices.