Multivariate causal effects: a Bayesian causal regression factor model

Area of Science:

• Environmental Science
• Public Health
• Statistical Modeling

Background:

• Wildfire smoke is a major contributor to air pollution, affecting public health through complex chemical mixtures.
• Previous research primarily linked wildfire smoke to total particulate matter (pm$_{2.5}$), leaving the causal impact on specific chemical compositions understudied.

Purpose of the Study:

• To investigate the causal relationship between wildfire smoke and the chemical composition of pm$_{2.5}$.
• To estimate the multivariate causal effects of wildfire smoke on 27 chemical species concentrations in pm$_{2.5}$ across the United States.

Main Methods:

• Developed a Bayesian causal regression factor model incorporating a causal inference framework for multivariate potential outcomes.
• Introduced a novel Bayesian factor model using a probit stick-breaking process prior for treatment-specific factor scores.
• Addressed missing data challenges and characterized latent factor structures crucial for multivariate outcome correlations.

Main Results:

• Monte Carlo simulations confirmed the model's accuracy in estimating causal effects and latent structures.
• Applied the model to US air quality data, revealing causal impacts of wildfire smoke on 27 pm$_{2.5}$ chemical species.
• Provided insights into the interdependencies between wildfire smoke exposure and specific chemical components of pm$_{2.5}$.

Conclusions:

• The developed Bayesian causal regression factor model effectively quantifies the multivariate causal effects of wildfire smoke on pm$_{2.5}$ chemical composition.
• This research enhances understanding of air quality impacts from wildfires, crucial for public health assessments.
• The methodology offers a flexible approach to analyzing complex, multivariate environmental data with missing observations.