Propensity score weighting analysis and treatment effect discovery.

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Propensity score analysis is crucial for estimating causal effects from observational data.
• Inverse probability weighting (IPW) is a common method, but can suffer from instability due to large weights when treatment group overlap is limited.
• This instability can bias estimates of treatment effects and their variance.

Purpose of the Study:

• To introduce and evaluate a class of modified inverse probability weighting (IPW) estimators designed to mitigate issues arising from poor overlap in propensity score distributions.
• To provide theoretical justification for deviations from the standard average treatment effect estimand.
• To demonstrate substantial improvements in statistical power and estimation stability compared to existing methods.

Main Methods:

• Development of modified inverse probability weighting (IPW) estimators.
• Analytical variance estimation accounting for propensity score sampling variability.
• Augmentation with outcome models to enhance efficiency, mimicking double robustness.
• Extensive simulations and real-world data application for validation.

Main Results:

• Modified IPW estimators effectively address numerical instability caused by large weights in low-overlap scenarios.
• The modified approach achieves significant gains in statistical power compared to standard IPW and other propensity score methods.
• Analytical variance estimates provide accurate adjustments for estimated propensity scores.
• The methodology demonstrates improved efficiency when combined with outcome modeling.

Conclusions:

• Modified inverse probability weighting offers a robust solution for propensity score analysis with limited treatment group overlap.
• The proposed methods enhance statistical power and estimation stability, crucial for reliable causal inference.
• The R package PSW implements these advanced techniques for practical application.