A multiple imputation strategy for incomplete longitudinal data

Longitudinal studies are commonly used to study processes of change. Becaus e data are collected over time, missing data are pervasive in longitudinal studies, and complete ascertainment of all variables is rare. In this paper a new imputation strategy for completing longitudinal data sets is propose d. The proposed methodology makes use of shrinkage estimators for pooling i nformation across geographic entities, and of model averaging for pooling p redictions across different statistical models. Bayes factors are used to c ompute weights (probabilities) for a set of models considered to be reasona ble for at least some of the units for which imputations must be produced, imputations are produced by draws from the predictive distributions of the missing data, and multiple imputations are used to better reflect selected sources of uncertainty in the imputation process. The imputation strategy i s developed within the context of an application to completing incomplete l ongitudinal variables in the so-called Area Resource File. The proposed pro cedure is compared with several other imputation procedures in terms of inf erences derived with the imputations, and the proposed methodology is demon strated to provide valid estimates of model parameters when the completed d ata are analysed. Extensions to other missing data problems in longitudinal studies are straightforward so long as the missing data mechanism can be a ssumed to be ignorable. Copyright (C) 2001 John Wiley & Sons, Ltd.