Panpan Zhang
Statistical Analysis of Incomplete Longitudinal Data Under Different Missing Scenarios
Presenter
![Photo of Panpan Zhang](http://dbei.nmsdev3.com/sites/default/files/uploads/styles/width470/public/PanpanZhang.jpg?itok=LoBaDCtn)
I have a postdoctoral researcher under Professor Sharon Xie's supervision. Our primary reserach focus is developing novel statistical methods for missing data and longitudinal data motivated by neurodegenerative diseases. Besides, I have interest in network analysis and probablistic graphical models.
Abstract
In this research, we compare the performance of available-case analysis (ACA) and multiple imputation (MI) dealing with missing data in longitudinal studies. More precisely, we consider missing data occurring in longitudinal responses, covariates or both under different scenarios of missingness specified by missing data generation procedure. The missing data may depend on observed responses only, fully observed covariates, partially observed covariates, or a combination of observed responses and covariates, etc. For each scenario, extensive simulations are carried out, where the performance of ACA and MI is assessed through estimation bias and relative efficiency. All the simulations are done under the assumption of missing at random (MAR). In addition, we apply the methods to a neurodegenerative disease data set.
Keywords
Longitudinal data; missing data; multiple imputationCommenting is now closed.
About Us
To understand health and disease today, we need new thinking and novel science —the kind we create when multiple disciplines work together from the ground up. That is why this department has put forward a bold vision in population-health science: a single academic home for biostatistics, epidemiology and informatics.
© 2023 Trustees of the University of Pennsylvania. All rights reserved.. | Disclaimer
Comments
Thank you for the questions. In the simulations, the missing outcomes do not have to be monotone (e.g. missing the second visit but coming back in the next). In fact, in the real data example, we have observed intermittent pattern in the missing data, too.
The assumption of MAR is primarily based on the collaborators' expertise and research experience. To the best of my understanding, there is no evidence to show that the longitudinal outcomes are not missing at random. However, this assumption has not yet been rigorously justified. We will investigate that and get back to you.