|
Activity Number:
|
625
- Environmental Epidemiology and Spatial Statistics
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 3, 2017 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #324556
|
|
|
Title:
|
Dynamic Spatial-Temporal Point Process Models via Conditioning
|
|
Author(s):
|
Athanasios Micheas* and Justin Okenye and Christopher Wikle
|
|
Companies:
|
Univ of Missouri- Columbia and University of Missouri and University of Missouri
|
|
Keywords:
|
Data Augmentation ;
Marked Hawkes Process ;
Finite Mixture Models ;
Marked Non-Homogeneous Poisson Point Process
|
|
Abstract:
|
We propose a Dynamic Spatial-Temporal Poisson Point Process (DSTPPP) model that accounts for the temporal and spatial clustering as a Marked Non Homogeneous Poisson Point Process (MNHPPP) via conditioning. We propose models for the intensity function of a (DSTPPP) using a Marked Non-Homogeneous Poisson Point Process (MNHPPP) via conditioning arguments that allow for additional interpretations. More precisely, the DSTPPP intensity function is modeled as a product of two functions; the first one representing the intensity of the temporal (time) component and the second a function of the intensity of the spatial (location) process, which is defined conditionally upon the temporal component of the event. In particular, the temporal component is modeled by a Hawkes Process and the spatial (location) process is modeled via a mixture of bivariate normal components. We propose a flexible hierarchical Bayesian framework for estimation of the parameters of the intensity function using a Markov chain Monte Carlo (MCMC) algorithm. The methodology is illustrated using simulated data and an application involving modeling and inference for earthquake data.
|
Authors who are presenting talks have a * after their name.
