Copula-Based Method Corrects Endogeneity in Treatment Effect Estimation
A novel statistical technique tackles unobserved confounding in the doubly robust estimation of treatment effects, especially significant in healthcare studies where proxy variables such as prescription refill rates are endogenous. This copula-corrected doubly robust estimator employs Gaussian copulas to represent the joint distribution of endogenous covariates alongside the error term, removing the necessity for instrumental variables. It preserves the doubly robust characteristic, necessitating accurate specification of either the treatment or outcome model. The method is validated through Monte Carlo simulations, effectively addressing endogeneity in both models at the same time.
Key facts
- Doubly robust estimation assumes no unobserved confounding.
- Proxy variables like prescription refill rates are often endogenous.
- Copula-corrected estimator addresses endogeneity without instrumental variables.
- Gaussian copulas model joint distribution of endogenous covariates and error term.
- Method preserves doubly robust property.
- Monte Carlo simulations demonstrate effectiveness.
- Applicable to healthcare research with unmeasured confounding.
- Corrects endogeneity in both treatment and outcome models.
Entities
—