Spatial point process modeling

Point pattern data is a unique type of spatial data where the location of incidence is considered random. Point processes model the occurrence and location of a certain incidence over time or in a bounded region. My current research focuses on developing models for spatial point process that are both flexible and computationally efficient over irregular observation windows.

Modeling for Poisson process over irregular spatial domains

(with Anathasios Kottas)

• A Bayesian nonparametric prior for the Poisson process intensity function that respects the shape of the irregular domain
• A unified prior model for the intensity function and total intensity
• Flexible (implies a Bernstein-Dirichlet prior on the Poisson density function) and computationally efficient (efficient Gibbs updates in the posterior simulation)
• Draft is available here