Estimating the probability of New Zealand regions being free from COVID-19 using a stochastic SEIR model

Rafal Bogacz

To assess accuracy of this parameter estimation procedure we tested if the correct values of parameters can be recovered from artificially generated data. We generated 20 artificial datasets, each containing a similar amount of data that is available on the current history of cases in New Zealand. For each dataset we randomly generated γ from uniform distribution between 0.1 and 0.4, and generated β from uniform distribution between 0.05 and γ + 0.05. With these parameters we generated 20 sequences (corresponding to 20 DHB) of ∆R t . For each DHB, the initial numbers of presymptomatic and infectious individuals E 0 and I 0 were generated from uniform distribution between 1 and 10. The model was then simulated for 21 days according to Equations 1-5. For each dataset we estimated β and γ and compared with true values from which data were generated.