Title: Recent advances in iterative imputation models
Authors: Gerko Vink - Utrecht University (Netherlands) [presenting]
Stef van Buuren - University of Utrecht (Netherlands)
Abstract: People are nowadays aware that not treating missing data problems may be the least desirable approach to drawing inference from incomplete data. Many routines to handle missing values are therefore included in standard statistical software. Most of the incomplete data problems are unfortunately not standard and require approaches that go beyond the standard solutions: some problems would benefit from a multivariate imputation approach, while univariate imputation would be more flexible for other problems. Recent advances in iterative imputation are detailed that allow for closer modeling of the imputation problem at hand. The focus lies on hybrid imputation, where univariate and multivariate imputation methods are combined through the specification of imputation blocks. We will explore the theory and application of some of these newer imputation models.