A comparison of missing data handling methods in linear structural relationship model: evidence from BDHS2007 data

Abstract

Missing observations in dependent variable is a common feature in survey research. A number of techniques have been developed to impute missing data. In this article, we have evaluated the performance of several impu- tation methods namely mean-before method, mean-before-after method and expectation-maximization algorithm in linear structural relationship model. On the basis of mean absolute error and root mean square error for both simulated and real data sets, we have shown that expectation-maximization algorithm is the most effective method than the other two imputation meth- ods to analyze the missing data in linear structural relationship model.