In real epidemic processes, the basic reproduction number R0 is the combined outcome of multiple probabilistic events. Nevertheless, it is frequently modeled as a deterministic function of epidemiological variables. This paper discusses the importance of adequate treatment of uncertainties in such models. This is done by comparing two methods of uncertainty analysis: Monte Carlo uncertainty analysis (MCUA) and the Bayesian melding (BM) method. These methods are applied to a model for the determination of R0 of dengue fever based on entomological parameters. The BM was shown to provide a complete treatment of the uncertainties associated with model parameters. In contrast to MCUA, the incorporation of uncertainties led to realistic posterior distributions for parameter and variables. The incorporation, by the BM, of all the available information, from observational data to expert opinions, allows for the constructive use of uncertainties generating informative posterior distributions for all of the model's components that are coherent as a set.
Bayes Theorem; Dengue; Epidemiologic Models; Uncertainty
