Relaxed Doubly Robust Estimation in Causal Inference.

Area of Science:

• Causal inference
• Statistical modeling
• Observational studies

Background:

• Causal inference is vital in biomedical and social sciences.
• Doubly robust estimators offer consistency if either the propensity score or outcome model is correct.
• Semiparametric models balance interpretability and adaptability.

Purpose of the Study:

• Introduce a novel relaxed doubly robust estimator.
• Reduce the requirement for full model specification in causal inference.
• Enhance flexibility in semiparametric causal inference.

Main Methods:

• Developed a relaxed doubly robust estimator.
• Focused on semiparametric models for propensity score and outcome mean.
• Analyzed estimator's double robustness and semiparametric efficiency.
• Conducted simulation studies.

Main Results:

• The proposed estimator requires only consistent parameter estimation, not correct function specification.
• Demonstrated double robustness and semiparametric efficiency.
• Simulation studies validated practical implications.

Conclusions:

• The relaxed doubly robust estimator offers a more flexible approach to causal inference.
• Partially correct model specification is sufficient for valid inference.
• The method has practical utility in observational studies.