Table 1.

Model Fits to Early Time Data

(Left) Fit of confirmed case (conf.) or mortality (mort.) data from California (CA), Indiana (IN), and New York (NY) states to three different models (SIR, SEIR, and branching process) using Poisson regression. The table shows the log likelihood and the relative likelihood exp($\left(AI{C}_{min}-AIC\right)=2$) based on the Akaike information criteria (AIC) (39). The blue lettering corresponds to the lowest AIC value. The branching process parameters include a Weibull shape $k$ and scale $b$ for $w\left(t\right)$, along with the exogenous rate $\mu$. The table also shows parameters from the fit and the projected date for the peak in new cases for each of these datasets; the projected peak date for the branching process is made using the HawkesN model. For each state, we run the fit on both confirmed case data and mortality data, taken from ref. 17. (Right) Shown are the actual data points compared with the fitted curves.