Description
There is growing interest in conducting large-scale association studies in real-world observational data. One of the primary challenges is addressing extensive missing data. While imputation via multiple imputation by chained equation is a common approach, it can be computationally slow or infeasible in data with large number of variables. We propose an alternative procedure computes imputations in parallel. The procedure achieves comparable performance to sequential imputations and offers substantial speed up in computation time.
Learning Objective: Learn challenges and possible solutions in imputing extensive and mixture types of missing data in electronic health record systems.
Authors:
Zeling He (Presenter)
Harvard School of Public Health
David Cheng, VA Boston Healthcare System
Presentation Materials: