A Tutorial for Propensity Score Weighting for Moderation Analysis With Categorical Variables: An Application...

Area of Science:

• Epidemiology
• Biostatistics
• Social Sciences Research Methods

Background:

• Observational studies often face challenges with covariate imbalance across treatment groups, potentially biasing moderation effect estimates.
• Moderation analysis is crucial for understanding how relationships between variables differ across subgroups.

Purpose of the Study:

• To provide a step-by-step guide and code (STATA/R) for employing propensity score (PS) weighting in moderation analyses with categorical variables.
• To illustrate the application of PS weighting to address covariate imbalance within moderator subgroups.

Main Methods:

• The tutorial outlines key steps: examining covariate overlap, estimating PS weights within moderator levels, assessing covariate balance post-weighting, estimating moderated effects, and conducting sensitivity analyses.
• A case study utilized data from 41,832 adults (2019 National Survey on Drug Use and Health) to test gender as a moderator of the association between sexual minority status and adult smoking prevalence.

Main Results:

• Propensity score (PS) weights successfully achieved covariate balance within both gender groups in the case study.
• PS-weighted analyses revealed significant evidence of moderation, with sensitivity analyses indicating robustness for one gender group but not the other.

Conclusions:

• Propensity score (PS) weighting within moderator levels can effectively minimize bias caused by covariate imbalances.
• This method enhances the accuracy of estimated moderation effects in observational research.