Multiple imputation in three-level data structures


Multilevel data with three levels of hierarchy are common in health research studies. A common problem in such studies is the presence of missing data and often handled with multiple Imputation (MI). To our knowledge there are only two MI implementations that are specialized for imputing missing data in a three-level setting (one within R and the other in the stand-alone software Blimp) and a lack of sufficient guidance for practitioners regarding the settings for which each of these approaches is appropriate. We investigate the performance of alternative MI approaches for handling three-level incomplete data by means of a simulation study under a number of different scenarios. Based on a case study from the Childhood to Adolescence Transition Study (CATS), we compared a range of currently available multilevel MI methods designed for single-level and two-level data combined with ad hoc approaches, such as the use of dummy indicators (DI) for school clusters or a just another variable (JAV) approach to repeated measures, in terms of bias and precision.

Jul 20, 2019 12:00 AM
Theatre 1, Level B1, 207-221 Bouverie St · Carlton