457 – Missing and Interval-Censored Data
A Simulation Study to Compare Missing Data Imputation Methods for Binary Outcome Variables Binary Outcome Variables
Fang Liu
Merck
Chen Jingjing
MedImmune
The analysis of longitudinal data in clinical trials presents a challenge as there are often missing data points. When binary outcomes variables are involved, the missing data imputation methods may become complicated. A simulation study illustrates how Generalized Linear Mixed Model (GLMM), Inverse Probability Weighted (IPW) Generalized Estimation Equation (GEE) method, multiple imputation and doubly robust method work in practice, especially for binary outcome variables in terms of efficiency and accuracy, with MAR assumption.