Abstract for: An Exploratory Evaluation into the Effect of Data Availability and Indicator Coverage on Parameter Estimates using HMC

There are typically two important questions addressed via the model calibration process: (1) does the time series of the fitted model match to the historical data; and (2) can reliable parameter estimates be inferred that are bounded within credible intervals. The evolution of Markov Chain Monte Carlo (MCMC) methods provide powerful methodological and computational frameworks for parameter estimation, and recent studies confirm the value of the Hamiltonian Monte Carlo approach for system dynamics models. This paper addresses an important research question for the calibration process, namely: what is the impact of data availability and indicator coverage on parameter estimation. It presents a 3 x 7 factorial study based on an SEIR model with cases, hospitalisations, and deaths. An exploratory analysis is presented, where all models converge. Our results highlight differences for a number of posterior distributions calculated, depending on the data availablity, and the set of indicators used.