An epidemiological forecast model and software assessing interventions on COVID-19 epidemic in China

Peter X Song; Lili Wang; Yiwang Zhou; Jie He; Bin Zhu; Fei Wang; Lu Tang; Marisa Eisenberg

First, we show the analysis of the Hubei COVID-19 data. Note that option dic=T enables to calculate the deviance information criterion (DIC) for model selection, while options, save_files=T and save_mcmc, allow the storage of MCMC output tables, plots, summary statistics and even full MCMC draws, which may be saved via the path of file_add, or otherwise via the current working directory. The major results returned from the package include predicted cumulative proportions, predicted turning points of interest, two ggplot2 [36] objects of the summary plots related to both infection and removed compartments, a summary output table containing all the posterior means, median and credible intervals of the model parameters, and DIC if opted. The traceplots and other useful diagnostic plots are also provided and automatically saved if save_files=T is opted. In the package, function tvt.eSIR() works on the epidemiological model with timevarying transmission rate in Section 2.2, and qh.eSIR() for the other epidemiological model with a quarantine compartment in Section 2.3. Note that in function tvt.eSIR(), with a choice of exponential=F, a step function is run in the MCMC when both change_time and pi0 are specified. To fit the model with a continuous transmission rate modifier function, user may set exponential=T and specify a value of lambda0. The default is to run the basic epidemiological model with no quarantine or πptq " 1 in Section 2.1. death_in_R is usually set to be the average ratio of death and removed proportions at each observation time point, which is used to estimate the death curve in the forecast plot of the removed compartment. Below are the R scripts used in the analysis.