Propensity score stratification methods for continuous treatments.

Area of Science:

• Causal inference methods
• Observational data analysis
• Epidemiology

Background:

• Propensity score methods are crucial for causal inference in observational studies but are predominantly developed for binary treatments.
• Existing propensity score methods for continuous treatments, primarily weighting techniques, can suffer from poor covariate balance and unstable estimates due to extreme weights.

Purpose of the Study:

• To introduce and evaluate novel propensity score stratification techniques for continuous treatments.
• To address the limitations of current weighting methods in continuous treatment settings.

Main Methods:

• Development of generalized propensity score cumulative distribution function (GPS-CDF) and nonparametric GPS-CDF approaches.
• Utilizing empirical cumulative distribution functions (ECDFs) for stratifying subjects based on pretreatment confounders.
• Comparison with standard weighting techniques through a detailed simulation study.

Main Results:

• The proposed empirical CDF-based stratification methods demonstrated superiority over standard weighting techniques in simulations.
• These new methods provide improved covariate balance and more stable causal estimates compared to weighting.
• Application to the "Mexican-American Tobacco use in Children" study revealed a causal link between movie smoking imagery exposure and adolescent smoking behavior.

Conclusions:

• Novel propensity score stratification methods based on empirical CDFs offer a robust alternative to weighting for continuous treatments.
• These techniques enhance the reliability and accuracy of causal effect estimation in observational research.
• The findings provide valuable new tools for researchers investigating continuous exposures and their causal impacts.