Multiple Imputation of Missing Multilevel, Longitudinal Data: A Case When Practical Considerations Trump “Best Practices”?

• In Event: Issues in Clustered Data Structures

Abstract

Despite advances in the field of missing data analysis, there exists uncertainty about how best to handle missing multilevel, longitudinal data. We propose that the implementation of multiple imputation – even though it is not necessarily designed to handle complex data – to fill-in missing multilevel, longitudinal data may be a specific case of when educational researchers must temporarily put aside issues of data analytic “best practises” in favour of practical considerations. We provide readers with a resource to which to refer when dealing with incomplete multilevel, longitudinal data in their own research contexts -- by providing a review of the missing data literature, and by providing a step-by-step explanation of the syntax used to multiply-impute an illustrative data set.

Authors

• Jennifer E.V. Lloyd, The University of British Columbia

• Jelena Obradović, Stanford University

• Richard M. Carpiano, The University of British Columbia

• Frosso Motti-Stefanidi, University of Athens