Evaluating two statistical methods for handling dropout impact in repeated measures data: A comparative study on missing at random assumption

Mohyaldein Salih^1*Ali Satty²Henry Mwambi¹

• ¹University of KwaZulu-Natal, Durban, South Africa
• ²Northern Border University, Arar, Saudi Arabia

This paper focuses on analyzing repeated measures data in the presence of dropouts. The primary aim is to examine and compare the effectiveness of multiple imputation (MI) and inverse probability weighting (IPW) methods for addressing dropout missingness under the missing at random (MAR) assumption. The comparison is based on repeated measures data generated from a marginal regression model, considering diverse sample sizes and varying dropout rates. The methods are evaluated using three performance metrics: bias, coverage, and mean square error (MSE). Results from the simulation indicate that MI consistently outperforms IPW, yielding lower bias and MSE, and higher coverage, particularly as dropout rates increase. An empirical evaluation using clinical trial data on serum cholesterol further supports the superior stability and precision of MI estimates, despite both methods producing similar inferential conclusions. These findings suggest that MI offers a more reliable approach for handling dropout in repeated measures studies, especially when dropout is moderate to high and the MAR assumption is plausible.