Use of Stochastic Volatility Models in Epidemiological Data: Application to a Dengue Time Series in São Paulo City, Brazil

Abstract

Background: A study on the dengue daily counting in São Paulo city in a fixed period of time is assumed considering a new regression model approch.

Methods: Under a Bayesian approach, it is introduced a polynomial linear regression model in presence of some covariates which could affect the counts of dengue in São Paulo city considered in the logarithm scale, combined with existing stochastic volatility models usually assumed in financial data analysis. Markov Chain Monte Carlo methods are used to get the posterior summaries of interest.

Results: The new model approach showed some advantages when compared to other existing times series models usually used to model epidemics data.

Conclusion: The use of the polynomial regression model combined with existing volatility models under a Bayesian approach showed that it is possible to get very accurate fit for the counting dengue data in São Paulo city where it is possible to jointly model the means and volatilities (variances) of the epidemiological dengue time series.