DefinePK

Abstract

Missing data is a common problem faced by researchers in many studies. The occurrence of missing data can produce biased results at the end of the study and affect the accuracy of the findings. There are various techniques to overcome this problem and multiple imputation technique is the best solution. Multiple imputation can provide a valid variance estimation and easy to implement. This technique can produce unbiased result and known as a very flexible, sophisticated approach and powerful technique for handling missing data problems. One of the advantages of Multiple Imputation is it can use any statistical model to impute missing data. Hence the selection of the imputation model must be done properly to ensure the quality of imputation values. However the selection of imputed model is actually the critical step in Multiple Imputation. This research study a linear regression model (LR) as the selected imputation model, and proposed the new algorithm named Linear Regression with Half Values of Random Error (LReHalf). The proposed algorithm is used to improve the performance of linear regression in the application of Multiple Imputation. Furthermore this research makes comparison between LR and LReHalf. The performance of LReHalf is measured by the accuracy of imputed data produced during the experiments. Future research is highly suggested to increase the performance of LReHalf model. LReHalf was recommended to enhance the quality of MI in handling missing data problems, and hopefully this model will benefits all researchers from time to time.