Two-step estimation in ratio-of-mediator-probability weighted causal mediation analysis

Area of Science:

• Causal Inference
• Statistical Modeling
• Epidemiology

Background:

• Causal mediation analysis decomposes treatment effects into indirect and direct pathways.
• Propensity score-based weighting is used to adjust for confounding in observational studies.
• Standard error estimation in two-step procedures can be problematic due to ignoring weight estimation uncertainty.

Purpose of the Study:

• To investigate appropriate estimation of estimator variability in propensity score-weighted causal mediation analysis.
• To extend a score function stacking solution to ratio-of-mediator-probability weighting.
• To provide accurate standard error estimation for indirect and direct effects.

Main Methods:

• Developed a two-step estimation procedure using propensity score-based weighting.
• Employed ratio-of-mediator-probability weighting to estimate causal effects.
• Utilized a score function stacking approach to derive the asymptotic variance-covariance matrix.

Main Results:

• Simulation results demonstrate that sampling uncertainty in estimated weights is significant and should not be ignored.
• The proposed stacking procedure provides a viable alternative to bootstrap standard error estimation.
• The method yields accurate standard errors for indirect and direct effect estimators.

Conclusions:

• The score function stacking method offers a robust approach for standard error estimation in propensity score-weighted causal mediation analysis.
• Ignoring weight uncertainty can lead to incorrect statistical inferences.
• This approach has broad implications for causal analyses employing propensity score-based weighting.