Causal inference with incomplete compliance or outcomes truncated by death
Principal Investigator
Name
Aixian Chen
Degrees
Ph.D student
Institution
School of Economics and Statistics, Guangzhou University, Guangzhou, China
Position Title
Ph.D student
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-1069
Initial CDAS Request Approval
Oct 11, 2022
Title
Causal inference with incomplete compliance or outcomes truncated by death
Summary
In recent years, causal inference has attracted extensive attention in both clinical medicine and human society. Both truncated by death and incomplete compliance are common problems in clinical medicine. This project is devoted to the following two aspects: First, when the outcome variable is truncated by death, in which patients die before outcomes of interest are measured. We can construct the confidence interval of partially identifiable model or solve its boundary by linear programming method, or even realize the point identification of SACE by adding assumptions. For example, under the principal stratification framework proposed by Frangakis and Rubin, a surrogate variable is introduced to distinguish the mixed distribution functions of two potential layers. Second, when individuals do not fully comply with the random assignment, that is, there is a problem of sample selection, which will lead to sample bias if directly ignored. We can introduce effective instrumental variables to identify the Complier Average Causal Effect (CACE). Alternatively, semi-parametric or non-parametric models can be further established based on Heckman Selection Model. Finally, the relevant simulation research and actual data analysis are carried out.
Aims
1. Bounding on the causal effect with outcomes truncated by death, that is, studying the extent to which receiving a particular treatment reduces mortality；
2. Estimating the Treatment Effects from Studies with Imperfect Compliance.
Collaborators
Cui Xia, Guangzhou University.