Evento adiado para o dia 26 de janeiro de 2022.

Por Nuno Sepúlveda (Faculty of Mathematics & Information Science, Warsaw University of Technology, Poland & CEAUL - Centro de Estatística e Aplicações da Universidade de Lisboa, Portugal).

Serological data analysis has often the objective of estimating the seroprevalence, i.e.,  the proportion of antibody-positive (i.e., seropositive) individuals in the population. However, serological data are intrinsically quantitative and, therefore, seropositive and seronegative individuals need to be estimated from the data. A simple approach to this estimation problem is to classify  individuals as seropositive if their antibody values exceed a certain threshold; otherwise, they are considered as antibody-negative (or seronegative). The threshold for antibody positivity is routinely determined by the three-sigma rules based on extreme quantiles of the normal distribution. In this rule, the threshold for seropositivity is estimated as the mean plus three standard deviations from the estimated antibody distribution of the seronegative population. In this talk, I will discuss two inferential problems - estimation bias and apparent control of specificity - arising from this rule. These problems will be discussed with public data on serological testing against the SARS-CoV2. In the end, one can ask the question: is the three-sigma rule a beautiful statistical construct or rather a little beast hidden in serological data analysis?

Transmissão em direto via Zoom.