Causal Estimators in Meta-analysis with Missing Data
Principal Investigator
Name
Ruofan Bie
Degrees
Ph.D
Institution
Brown University
Position Title
Student
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-1146
Initial CDAS Request Approval
Oct 24, 2023
Title
Causal Estimators in Meta-analysis with Missing Data
Summary
In this project, we proposed causal estimators to transport estimation of causal average treatment effect from trials to target in meta-analysis, while handling two types of missing data. In this project, we define trials as an individual participant data meta-analysis that studies the same treatment effect; and we define target as a target sample that collects the same covariates as the trials but lack randomization information. We proposed g-formula estimator, IPW estimator and doubly-robust estimator to identify the causal average treatment effect in the target using trial data information. We provided analytical proof of the identification and large sample property of the proposed causal estimators and used simulation study to show the advantage of the proposed estimators over multiple-imputation estimators and pooled estimators. In the next step, we plan to use real-data analysis to further verify the feasibility and performance of our proposed estimator on real data. We plan to use the NLST data as the target and try to transport from the PLCO data the estimation of causal average treatment effect on NLST data.
Aims
1. propose causal estimators for transportability in meta-analysis while handling two types of missing data (completed)
2. provide analytical proof of the identification and large sample property of the proposed estimators (completed)
3. run simulation study to compare proposed estimators with alternative methods (completed)
4. run real-data analysis to verify the feasibility of the proposed estimators (need NLST data)
Collaborators
Jon Steingrimsson, Brown University