Estimation of causal quantile effects with a binary instrumental variable and censored data.

Area of Science:

• Social, biological, and health sciences research.
• Epidemiology and biostatistics.
• Econometrics and causal inference.

Background:

• Determining causal treatment effects is crucial across scientific disciplines.
• Instrumental variable (IV) methods are vital for addressing unmeasured confounding.
• Existing methods face challenges with randomly censored outcomes.

Purpose of the Study:

• To propose a new framework for causal effect estimation using binary instrumental variables with censored outcomes.
• To introduce and define the complier quantile causal effect (CQCE) as a novel estimand.
• To establish conditions under which CQCE is identifiable and offers advantages over average causal effects in censored data scenarios.

Main Methods:

• Development of a novel binary instrumental variable framework.
• Quantification of causal effects via the complier quantile causal effect (CQCE).
• Adaptation of censored quantile regression with a derived weighting scheme for robust nonparametric estimation.
• Rigorous establishment of asymptotic properties for the proposed estimator.

Main Results:

• The proposed CQCE is identifiable under weaker conditions than traditional complier average causal effects with censored data.
• A practical weighting scheme is developed for stable estimation using existing software.
• Extensive simulations confirm the estimator's validity and satisfactory finite-sample performance.
• Application to a bone marrow transplant dataset demonstrates real-world utility.

Conclusions:

• The novel CQCE provides a valuable tool for estimating causal effects in the presence of unmeasured confounding and censored outcomes.
• The proposed estimation method is robust, statistically sound, and practically implementable.
• This approach enhances causal inference capabilities in observational studies within health and social sciences.