Extrapolation of guidance from non-longitudinal clustered settings to longitudinal data

Determine whether findings from simulation studies conducted in non-longitudinal clustered data settings on the performance of multiple imputation methods can be validly extrapolated to longitudinal data settings when evaluating joint modelling and fully conditional specification approaches.

Background

The paper summarizes available guidance on multiple imputation approaches for longitudinal data and compiles it into Table 3. For several analysis scenarios, the table indicates "No clear guidance" because dedicated longitudinal simulations are lacking.

The authors note that although some guidance exists from other multilevel or clustered contexts, it is uncertain whether those results apply to longitudinal settings. This creates an unresolved question about the validity of extrapolating evidence across settings, which affects method choice and interpretation for longitudinal analyses.