Overview
Scott L. Zeger (Johns Hopkins Bloomberg School of Public Health, Biostatistics) is a biostatistician known for generalized estimating equations (GEE) and longitudinal data methods. With Dominici and Samet, he developed the two-stage Bayesian hierarchical design for the NMMAPS 20-city air pollution–mortality study, in which Stage 1 semiparametric Poisson GAMs are fitted per city and their sufficient statistics pooled via a normal–inverse-Wishart hierarchy at Stage 2.
Key Contributions
- Generalized estimating equations (Liang–Zeger 1986; Zeger–Liang–Albert 1988): With Kung-Yee Liang, co-invented GEE for longitudinal/clustered data — a working-correlation quasi-score with robust sandwich variance — and, with Liang and Paul Albert, formalized the subject-specific vs. population-averaged distinction. See Generalized Estimating Equations and Zeger-Liang-Albert (1988).
- Co-developed the NMMAPS two-stage hierarchical framework (Dominici-Samet-Zeger 2000), contributing the statistical design connecting city-level GAM estimates to a second-stage Bayesian pooling model.
- Instrumental in the semiparametric Poisson GAM specification at Stage 1, including the choice of natural spline degrees of freedom for seasonal and meteorological adjustment.
Related