A Tutorial for Propensity Score Weighting Methods Under Violations of the Positivity Assumption.

Area of Science:

• Causal inference
• Biostatistics
• Epidemiology

Background:

• Conventional causal estimands like ATE, ATT, and ATC can be unidentifiable due to positivity assumption violations.
• Weighted counterparts (WATE, WATT, WATC) offer alternative, identifiable causal estimands.
• Propensity score (PS) methods are crucial for addressing confounding in observational studies.

Purpose of the Study:

• To provide a comprehensive review of recent advances in propensity score (PS) weighting methods.
• To offer practical guidance on selecting target estimands, implementing PS-weighted estimators, and conducting diagnostic assessments.
• To introduce the ChiPS R package for facilitating PS-weighted causal inference.

Main Methods:

• Review of recent propensity score (PS) weighting methodologies.
• Guidance on selecting primary target estimands and supplementary analyses.
• Implementation of PS-weighted estimators and post-weighting diagnostic procedures.
• Demonstration using extensive simulation studies and two real-world case studies.

Main Results:

• The study demonstrates the utility of PS weighting for obtaining identifiable causal effects.
• Simulation studies confirm the pertinence and performance of various PS-weighted estimators.
• Case studies illustrate the application of these methods in public health research.

Conclusions:

• Propensity score weighting provides a robust framework for causal inference when positivity is violated.
• The ChiPS R package enhances the practical application of these advanced methods.
• This approach enables valid estimation of treatment effects in observational data, contributing to internal validity.