Pre conference: ‘Advanced multilevel imputation’
Multiple imputation of multilevel data is challenging. In general, we need to consider the role of the variable in the multilevel model, the size and number of clusters, the reasons for the missingness, the complexity of the complete-data model, and the match between the complete-data and imputation models.
This workshop shows how multiple imputation of multilevel data can be done in R using a combination of the packages mice, lme4, micemd, mitml and miceadds. We discuss the relative strengths of the joint modeling and fully conditional specification frameworks, and provide solutions for progressively more complex multilevel models. We alternate theory lectures with hands-on practicals, where you apply the methodology using worked examples on your laptop.
It is an advanced-level workshop that assumes basic familiarity with multilevel models, multiple imputation and R.