Modelo de conglomerados para el análisis bayesiano de datos epidemiológicos en áreas pequeñas

  1. Flórez Lozano, Karen Cecilia
Supervised by:
  1. Ana Corberán Vallet Director
  2. José Domingo Bermúdez Edo Director

Defence university: Universitat de València

Fecha de defensa: 31 May 2016

Committee:
  1. Francisco Montes Suay Chair
  2. Virgilio Gómez Rubio Secretary
  3. María Dolores Ugarte Martínez Committee member
Department:
  1. STATISTICS AND

Type: Thesis

Teseo: 421494 DIALNET lock_openRODERIC editor

Abstract

The model presented here does not require from the start to define dependency or distance between neighbours, but exposes a formulation variables where risk allocation can capture different risk structures. So, it is an alternative approach where the relative risks of small areas are assigned to underlying risks. There is an unknown number of latent risks and small areas are assigned to these latent risk. Areas in the same class have the same risk but they do not necessarily share any border. Our proposal applies ideas of mixtures models, cluster detection and latent structure models. To illustrate the accuracy of our procedure, we show the main results obtained in a simulation study. We study the incidence of chickenpox in the city of Valencia in 2008.