Evaluation and development of multiple imputation algorithms for dealing with longitudinal data
Principal supervisor: Prof Julie A. Simpson
Co-supervisors: A/Prof Katherine J. Lee, Dr Alysha M. De Livera, Dr Margarita Moreno-Betancur
Missing data is a common occurrence in longitudinal studies, which involve multiple waves of data collection. Traditional multiple imputation (MI) methods (fully conditional specification (FCS) and multivariate normal imputation (MVNI)) treat repeated measurements of the same time-dependent variable as just another ‘distinct’ variable for imputation and therefore do not make the most of the longitudinal structure of the data. The recently implemented two-fold fully conditional specification (two-fold FCS) algorithm, which restricts the imputation of a time-dependent variable to time blocks where the imputation model includes measurements taken at the specified and adjacent times, takes into consideration the temporal ordering of the data. This PhD project will focus on the evaluation of MI methods available in standard statistical software for handling missing data in longitudinal studies with complex scenarios such as the existence of a time varying covariate with a nonlinear relationship with time, imputation of longitudinal categorical variables with restrictions on transitions over time, and complex survey designs. Guidance on handling missing data in such scenarios is limited in the statistical literature. The performance of the existing methods under each scenario will be assessed through simulation studies and case studies designed based on the Longitudinal Study of Australian Children (LSAC).
Scholarship and funding body:
Victorian International Research Scholarship (VIRS) [State Government of Australia]
Melbourne International Fee Remission Scholarship [The University of Melbourne]