• @kassem
    Kassem CC BY-NC-SA
    3
    @simplicissimus
    @colporteur
    @ericw
    3

    ’There is no absolute truth’: an infectious disease expert on #Covid-19, misinformation and ’bullshit’ | World news | The Guardian
    ▻https://www.theguardian.com/world/2020/apr/28/there-is-no-absolute-truth-an-infectious-disease-expert-on-covid-19-mis

    We’re all used to verbal #bullshit. We’re all used to campaign promises and weasel words, and we’re pretty good at seeing through that because we’ve had a lot of practice. But as the world has become increasingly quantified and the currency of arguments has become statistics, facts and figures and models and such, we’re increasingly confronted, even in the popular press, with numerical and statistical arguments. And this area’s really ripe for bullshit, because people don’t feel qualified to question information that’s given to them in quantitative form.

    #manipulation #chiffres

    Kassem CC BY-NC-SA
    • @simplicissimus
      Simplicissimus

      Pour ma part, s’il y a un truc à retenir, c’est ça :

      The other big piece is understanding the notion of positive predictive value and the way false positive and false negative error rates influence the estimate. And that depends on the incidence of infection in the population.

      If you have a test that has a 3% error rate, and the incidence in the population is below 3%, then most of the positives that you get are going to be false positives. And so you’re not going to get a very tight estimate about how many people have it. This has been a real problem with the Santa Clara study. From my read of the paper, their data is actually consistent with nobody being infected. A New York City study on the other hand showed 21% seropositive, so even if there has a 3% error rate, the majority of those positives have to be true positives.

      Je sais, je me répète un peu, mais ça fait plaisir de le voir écrit noir sur blanc. Et ça ne vaut pas seulement pour la biologie : la détection de signaux faibles ne peut pas se faire avec des tests.

      Simplicissimus
    Écrire un commentaire