[Probabilidad-Estadistica-Seminario] Martes 28 de Mayo, Seminario IESTA (SIESTA)
gillanes en cmat.edu.uy
gillanes en cmat.edu.uy
Mie Mayo 22 12:40:49 -03 2019
Distribuyo info del seminario del IESTA.
Saludos!
El próximo seminario organizado por el IESTA será el Martes 28 de Mayo a las 14:00 horas en la sala de reuniones (Eduardo Acevedo 1139). El seminario estará a cargo de Ignacio Alvarez-Castro
IESTA-FCEA, quién presentará:
An Approximate Bayesian Computation of a Poisson Markov Random Field Model for Crash Data
Abstract
We propose a Winsorized Poisson Markov Random Field (WP- MRF) model to analyze intersection crashes. The WP-MRF model introduces spatial dependence at the observations level and incorpo- rates different amounts of spatial dependence in different directions. Due to the doubly intractable nature of the WP-MRF model we uti- lize approximate Bayesian computation (ABC) to estimate parame- ters. We study the performance of the WP-MRF model and the ABC approach to estimation through a simulation study based on crash frequency data in three Iowa cities: Ankeny, Marshalltown, and Tipton. We propose a measure of crash risk for each individual intersec- tion, based on the posterior predictive distribution. Departments of Transportation may use this measure to identify intersections with high risk, as complementary to more commonly used indicators as traffic volume or crash history. In Marshalltown, we find the spa- tial dependency of the number of crashes is significative in both the North-South and East-West directions even after conditioning on co- variates. In all three cities, we found high risk intersections with low traffic volume or low crash history.
Recuerden que todas las presentaciones y el programa tentativo de las próximas charlas está disponible en https://github.com/natydasilva/SIESTA (https://github.com/natydasilva/SIESTA)
Se agradece la difusión y esperamos puedan participar.
Saludos,
Natalia
------------ próxima parte ------------
Se ha borrado un adjunto en formato HTML...
URL: <http://www.cmat.edu.uy/pipermail/seminario-probabilidad-estadistica/attachments/20190522/b12312f6/attachment.html>
Más información sobre la lista de distribución seminario-probabilidad-estadistica