Evaluating the Effectiveness of Multiple Imputation for Handling Missing at Random Data : An Applied Study

Abstract

This research aims to assess  the effectiveness of the Multiple Imputation  (MI) method for processing  randomly  missing data.  These missing data consider an essential challenge in statistical analysis because they influence on accuracy of outcomes and  reliability of conclusions. Hence, suitable methods are required to handle such missing observations compares to maintain the robust statistical inference. This study also compares effectiveness of the Multiple Imputation (MI) with Deletion method, which is considered one of the most commonly used traditional techniques for treating missing data. Both methods are applied to handle missing values and to test their impact on statistical results. The simulation strategy was  utilized using  MATLAB 24 to generate data sets  consisting of  specific percentages of missing entries. The study was applies containing the following four main economic factors: exchange rate, price of a barrel of oil, foreign exchange reserves, and inflation rate which represent performance of the Iraqi economy during 2005-2024. The outcomes of the statistical analysis showed that both of Multiple Imputation Method and the Deletion Method provided statistically major results at a significance level of less than 0.05. This refers to that both methods persisted statistically significance even with presence of missing values. The findings also explained MI method obviously outperforms the Deletion method in retaining the true variance of the data and upholding the original sample size. Also it plays a major role in reducing bias and addressing  uncertainty, related to missing data. In particular if the percentage for missing observations is high. In addition, the results indicates that MI is more stable and realistic than the deletion method. In spite of the Deletion approach explained more than 86% of the model while the multiple imputation method explained less than 58%, nearly 28% of the explanations evaluated by the deletion method are misleading, primarily because of the loss some of original data. Thus, the results recommend that MI represents a suitable method  than the traditional deletion method for handling missing data.