A Comparison of The Methods of Estimating Missing Data Between Nearest Neighbor and Regression Imputation

Abstract

Missing data are observations that we need to know but cannot acquire. This is a problem that is often found in survey researches. Researchers must find a way to solve this data analysis problem. If the missing data are handled by an inappropriate method, the result of the analysis might be distorted and, consequently, the analysis of the rest of valid data could be deviated. This research studies 2 methods of solving the problem of missing data, which are Nearest Neighbor Imputation and Regression Imputation using the criterion of the root mean square error (RMSE) in the comparison of errors. It is found out from the comparison that one method may be appropriate for particular data or things that need to be considered only. However, the method of estimating missing data by Nearest Neighbor Imputation (NNI) is the appropriate means when considering from the criterion of the root mean square error (RMSE) since it yields minimum values in all levels of missing values.
The sample size equals 40.