A comparison of model‐ and multiple imputation‐based approaches to longitudinal analyses with partial missingness

Abstract

Longitudinal data sets typically suffer from attrition and other forms of missing data. Omissions, attrition, and planned missingness have limited our ability to conduct the most appropriate analyses. When this common problem occurs, several researchers have demonstrated that correct estimation with missing data can be obtained under mild assumptions concerning the missing data mechanism. An example application of latent growth curve methodology, analyzing the longitudinal developmental change in adolescent alcohol consumption, is presented (N = 586; 250 young men and 336 young women, 14 to 16 years of age at the initial time point). The analyses are conducted within a cohort‐sequential design, incorporating missingness introduced by design and due to attrition. We describe and illustrate 3 approaches to the analysis of missing data when some data are missing: multiple‐sample structural equation modeling procedures, raw maximum likelihood analyses, and multiple modeling and data augmentation algorithms.