2020/06/29 by Panos Toulis, Toulis, Panos
Medicine · #Applications (stat.AP) #COVID-19 Clinical Research Studies #Econometrics (econ.EM) #FOS: Computer and information sciences #FOS: Economics and business #Methodology (stat.ME) #SARS-CoV-2 and COVID-19 Research #SARS-CoV-2 detection and testing
paper · pdf · doi:10.48550/arxiv.2006.16214
openalex publication_date 2020/06/29 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
We propose a partial identification method for estimating disease prevalence\nfrom serology studies. Our data are results from antibody tests in some\npopulation sample, where the test parameters, such as the true/false positive\nrates, are unknown. Our method scans the entire parameter space, and rejects\nparameter values using the joint data density as the test statistic. The\nproposed method is conservative for marginal inference, in general, but its key\nadvantage over more standard approaches is that it is valid in finite samples\neven when the underlying model is not point identified. Moreover, our method\nrequires only independence of serology test results, and does not rely on\nasymptotic arguments, normality assumptions, or other approximations. We use\nrecent Covid-19 serology studies in the US, and show that the parameter\nconfidence set is generally wide, and cannot support definite conclusions.\nSpecifically, recent serology studies from California suggest a prevalence\nanywhere in the range 0%-2% (at the time of study), and are therefore\ninconclusive. However, this range could be narrowed down to 0.7%-1.5% if the\nactual false positive rate of the antibody test was indeed near its empirical\nestimate (~0.5%). In another study from New York state, Covid-19 prevalence is\nconfidently estimated in the range 13%-17% in mid-April of 2020, which also\nsuggests significant geographic variation in Covid-19 exposure across the US.\nCombining all datasets yields a 5%-8% prevalence range. Our results overall\nsuggest that serology testing on a massive scale can give crucial information\nfor future policy design, even when such tests are imperfect and their\nparameters unknown.\n