Lies, Gosh Darn Lies, and Not Enough Good Statistics: Why Epidemic Model Parameter Estimation Fails

Daniel E Platt; Laxmi E Parida; Pierre Zalloua

is the (which was not peer-reviewed) The copyright holder for this preprint . Figure 4 shows the evolution of the system variables in a linear-linear plot. The lags in the peak variables shown in figure 4a identify the peak pulse through the system of linear equations. The "est" entries in Table 1 for represent values commensurate with (but not a fit to) the NY hospitalization levels 21 . They are a factor of 12 smaller than those fitting the Wuhan hospitalization rate 8 . As such, it is clear that the impact of COVID on features such as progression to hospitalization, response to treatment for symptomatic patients, whether patients are identified in time to stop progression to serious or critical stages may impact survivability. The model predicts 3294 fatalities per million, peak recovering hospitalizations of 3347 on day 111, and peak mortality hospitalization (primarily long-term ventilator load) of 1732 on day 114. Figure 4b includes susceptible and recovered variables. The range of variation of these variables appears to dwarf the fraction of the population that is incubating, infected, or involved with hospital load. One feature of the equations is that the rate of flow of individuals through a compartment may not be reflected in the total number in the compartments at any given time, even at their peaks. At the end, these rates would leave 24,738 per million uninfected and susceptible, with 971,967 recovered per million. Table 2 . Exponential growth rates, corresponding doubling times for various populations and measurements given available data.