Raimund Bürger, Gerardo Chowell, Ilja Kröker, Leidy Y. Lara-Diaz:
Sensitivity and identifiability analysis for a model of COVID-19 in Chile
A compartmental model is formulated to describe the progression of the COVID-19 pandemic in Chile, where each of the 16 administrative regions is considered as a separate population. Parameters of this model, and in particular the basic reproductive number $R_0$, can be estimated by fitting the model to published information on the progression of the epidemic in each region. The adjustment of appropriate model parameters can be achieved by either the Simulated Annealing (SA) method or alternatively, by a stochastic optimization model using a classical Markov chain Monte Carlo (MCMC) technique. This estimation allows one to analyze the identifiability and sensitivity of the parameters. The approach considers the control policies applied in Chile for mitigating the outbreak of COVID-19, in particular the methodology assumed by the government to declare the quarantine measures. The outbreaks are followed in a regional scale taking into account the different dates of the first case, the quarantine measures adapted to the regional situation, and the different criteria applied to the selection of data.