Modeling the Drop-Out Mechanism in Repeated-Measures Studies

TLDR

Longitudinal studies frequently suffer from premature dropout, producing unbalanced data that modern software can handle but still yields biased inferences when plausible dropout mechanisms are not accounted for. This article proposes a unified model‑based framework that simultaneously models the longitudinal data and the dropout process, and recommends collecting dropout information to inform model selection. The framework classifies models into random‑coefficient selection and pattern‑mixture types, uses likelihood‑based inference (maximum likelihood or Bayesian), and includes examples, extensions, and sensitivity analyses when the dropout mechanism is uncertain. Keywords: Attrition, longitudinal data, missing data, nonrandom nonresponse, selection bias.

Abstract

Abstract Subjects often drop out of longitudinal studies prematurely, yielding unbalanced data with unequal numbers of measures for each subject. Modern software programs for handling unbalanced longitudinal data improve on methods that discard the incomplete cases by including all the data, but also yield biased inferences under plausible models for the drop-out process. This article discusses methods that simultaneously model the data and the drop-out process within a unified model-based framework. Models are classified into two broad classes—random-coefficient selection models and random-coefficient pattern-mixture models—depending on how the joint distribution of the data and drop-out mechanism is factored. Inference is likelihood-based, via maximum likelihood or Bayesian methods. A number of examples in the literature are placed in this framework, and possible extensions outlined. Data collection on the nature of the drop-out process is advocated to guide the choice of model. In cases where the drop-out mechanism is not well understood, sensitivity analyses are suggested to assess the effect on inferences about target quantities of alternative assumptions about the drop-out process. Key Words: AttritionLongitudinal dataMissing dataNonrandom nonresponseSelection bias