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

author/funder, who has granted medRxiv a license to display the preprint in perpetuity. In the above R coding scripts, only very short MCMC chains are specified for the consideration of running time. In practice, it is recommended to set M=5e5 and nburnin=2e5 to achieve desirable burn-ins and yield stable MCMC draws. We tried the step function given by Panel C in Figure 3 with pi0=c(1,0.9,0.5,0.1), an exponential function given by Panel B in Figure 4 with rate lambda0=0.05, both of which were compared with the basic model with πptq " 1. The results of the three modifier functions obtained from the tvt.eSIR function are summarized in Figures 9-11 . In these forecasting plots of the infected and removed compartments (Panel A and Panel C), the black dots left to the blue vertical line denote the observed proportions of the infected and removed compartments on the last date of available observations or before. That is, the blue vertical marks time t 0 as defined in Section 3. The green and purple vertical lines denote the first and second turning points, respectively. The salmon color area denotes the 95% credible interval of the predicted proportions Y I t and Y R t after t 0 , respectively, while the cyan color area represents either the 95% credible intervals of the prevalence (θ I t ) and or that of the probability of removal (θ R t ) prior to time t 0 . The gray and red curves are the posterior mean and median curves. The black curve in the plot (Panel C) is the proportion of deaths estimated from a pre-specified ratio death_in_R. The middle Panel B displays important dynamic features of the infection via a spaghetti plot, in which 20 randomly selected MCMC draws of the first-order derivative of the posterior prevalence of infection, namely 9