Using negative controls to identify causal effects with invalid instrumental variables.

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Traditional causal effect identification relies on instrumental variables with strong, untestable assumptions like unconfoundedness and exclusion restriction.
• Violations of these assumptions limit the applicability of standard methods in real-world observational studies.

Purpose of the Study:

• To propose a novel strategy for identifying causal effects under violations of instrumental variable assumptions.
• To develop a robust and efficient estimator for the average treatment effect in the treated (ATT).

Main Methods:

• Leveraging a negative control population or outcome to relax strong instrumental variable assumptions.
• Utilizing subpopulations with degenerate exposure and a parallel trend condition for instrument-outcome association.
• Developing semiparametric efficiency theory for a general instrumental variable model.

Main Results:

• A multiply robust and locally efficient estimator for the average treatment effect in the treated (ATT) was derived.
• The proposed method demonstrates potential for causal effect identification even when standard assumptions are unmet.
• Simulation studies and analysis of the Life Span Study validated the utility of the developed estimators.

Conclusions:

• The negative control strategy provides a viable alternative for causal inference when instrumental variable assumptions are violated.
• The developed estimator offers enhanced robustness and efficiency for estimating treatment effects in observational data.
• This work advances causal inference methodologies with practical implications for epidemiological and biomedical research.