Accuracy Evaluation of the MissForest Missing Data Imputation Approach in Social Science Data Analysis

This research examines the accuracy of the missForest missing data imputation approach. In the theoretical part of the work, the reasons for the appearance of omissions in sociological data are analyzed, various classifications of omissions are considered, and the influence of the types of omissions on the possibility of subsequent data processing. The analysis of both the missForest approach and other approaches is also carried out, and their advantages and disadvantages are considered. In the empirical part of the study, a secondary regression analysis of the ESS 2014 database is performed after generating omissions of three types of randomness in the fractions of 10, 30, and 50 percent and then imputing the omissions using the missForest approach. Using a metric based on a comparison of the 95% confidence intervals of the regression coefficients obtained on the data without omissions and the data after imputation of the generated omissions, we learn the limits of the applicability of the missForest approach, at what proportions of omissions in the data and at what types of the randomness of omissions we can apply the missForest approach without significant distortion of our analysis.