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

Daniel E Platt; Laxmi E Parida; Pierre Zalloua

The difficulty in understanding how the testing protocol impacts estimations of rates is illustrated in the New York State rates shown in Table 2 . Considering cases as a representative sample of a fixed proportion of the infected population argues for computing a rate based on cumulative cases. If, on the other hand, the testing generated a random sample of the broader population, more testing would identify more individuals simply because there were more tests. If so, the proportion of positives to total tests may be a closer approximation to the population, and the total positives would be proportional to the square of the actual proportion of diseases, resulting in a doubling of . That seems to be roughly what was observed between the two New York State regressions. On the other hand, cumulative rates for two other . CC-BY-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.