A comparative investigation of missing data handling methods in growth mixture modeling
Missing data can frequently appear in a longitudinal data analysis as they are measured repeatably. It is well known that missingness in empirical data can lead to biased estimation and invalid inference, and they should be treated properly. Researchers in behavioral sciences are often interested in modeling longitudinal developmental trajectories of individuals, such as the study of school achievement growth or self-regulation development. Growth Mixture Modeling (GMM) is statistical technique to identify unobserved subpopulations based on similar longitudinal growth trajectories. Due to common occurrence of missing data in longitudinal data, various methods are introduced to handle the missingness for GMM practice. For this dissertation project, the performance of missing data handling methodologies (Listwise deletion, Full Information Maximum Likelihood, and Multiple Imputation) was compared in terms of enumeration accuracy and relative bias through a series of Monte Carlo simulation. Based on the results of findings in this dissertation, practical guidelines for applied researchers are discussed.Embargo status: Restricted until 09/2172. To request the author grant access, click on the PDF link to the left.