Francesca Dominici

Overview

Francesca Dominici (Johns Hopkins Bloomberg School of Public Health, Biostatistics) is a biostatistician specializing in Bayesian hierarchical methods for environmental health research. Her 2000 paper with Samet and Zeger introduced a two-stage Bayesian hierarchical design for pooling city-specific air pollution–mortality associations across the 20 largest US cities, combining semiparametric Poisson GAMs at Stage 1 with a normal–inverse-Wishart hierarchy at Stage 2.

Key Contributions

• Developed the NMMAPS (National Morbidity, Mortality, and Air Pollution Study) two-stage hierarchical framework: city-specific Poisson GAMs feeding $\hat{\beta}_c, V_c$ into a Bayesian normal hierarchy with optional spatial correlation (Dominici-Samet-Zeger 2000).
• Demonstrated that a +10 μg/m³ increase in PM₁₀ is associated with a +0.48% increase in daily all-cause mortality (95% CI: 0.05–0.92) pooled across 20 US cities.
• Showed that city-level socioeconomic covariates fail to explain between-city heterogeneity ($\hat\sigma \approx 0.76$), and that the IW prior is sensitive with only 20 cities.

Related

• Bayesian Hierarchical Model
• Gibbs Sampler
• Metropolis-Hastings Algorithm
• Jonathan M. Samet
• Scott L. Zeger
• Dominici-Samet-Zeger (2000) Combining Evidence on Air Pollution and Daily Mortality from the 20 Largest US Cities