MODEL UNCERTAINTY IN THE COMPARISON OF TWO SINGLE DENGUE OUTBREAKS

Abstract

In recent years, there has been increased interest in using statistical models for analysis of single dengue outbreaks based on the reported cumulative cases. The three parameter logistic (3P logistic) and the Richards models have been used to estimate primary epidemiological parameters in single dengue outbreak. A topic that could be of interest to epidemiologists is the comparison of two single dengue outbreaks based on estimates of key epidemiological parameters: The turning point, the final
size and the basic reproductive number R0. In order to compare two single dengue outbreaks we create a model that takes into
account both outbreaks simultaneously. In this paper, we describe different methodologies based on Frequentist and Bayesian
approaches that takes into account the model uncertainty in the comparison of two single dengue outbreaks. The Frequentist
approach consists of comparing outbreak doing an extension of 3P logistic and Richards models and the use of model averaging
for taking into account model uncertainty. In the Bayesian approach, we use a Bayesian hierarchical model and we use Bayesian
model averaging applying Gibbs variable selection. The proposed methods are applied to dengue outbreaks that occurred in La
Lisa municipality, Havana City, Cuba during 2006 and 2007 outbreaks.