Parameter estimation for nonlinear disease dynamical system using particle filter
Muhammad Javvad ur Rehman, Sarat C. Dass, Vijanth Sagayan Asirvadam, Ahmed Adly · 2015
We address the issue of parameter estimation for nonlinear dynamical systems obtained as a model for dengue disease incidence. A Bayesian framework of estimation is adopted. Parameter estimation is performed using a Metropolis Hastings algorithm in which the target distribution of the resulting Markov chain equals the posterior distribution of unknown parameters. Intermediate predictive and filtering density evaluations required, within each Metropolis-Hastings step are evaluated using the particle filters (PF). The methodology is used to estimate unknown parameters governing the evolution of an underlying state space representing the dynamics of the force of infection. We illustrate our estimation methodology on dengue incidences collected from 2009 – 2014 for the district of Gombak in Selangor, Malaysia.