Overview
Sonia Petrone is a statistician (Bocconi University, Milan), known for Bayesian nonparametrics, exchangeability and predictive inference, and Bayesian methods for curve estimation.
Key Contributions / Features
- Discreteness of the Dirichlet process prior (Petrone-Raftery 1997): with Raftery, a caution that the DP's discreteness distorts partition inference and Bayes factors.
- Bernstein polynomial priors for density estimation, and work on partial exchangeability, predictive characterizations, and hierarchical/dependent nonparametric models.
Related