I think the modeling without priors occurs because of the failed surveillance. If there had been aggressive testing, contact tracing, and quarantining, the spread might be have been stopped, but, even if it were not, at least there would
be case history stories that could be put into an agent-based model. Are spread rates on the subway different than at churches and sports events or Mardi Gras? There was no competent effort to do that, so modelers like Murray’s team fish for explanatory
variables retrospectively. Such models could probably make precise predictions if millions of people had test kits arrive in the mail the first week, and Apple and Google coordinated to have them relay diagnosis information (via smartphones) to a central
repository.
From: Friam <[hidden email]> on behalf of "[hidden email]" <[hidden email]>
Reply-To: The Friday Morning Applied Complexity Coffee Group <[hidden email]>
Date: Sunday, April 26, 2020 at 11:33 AM
To: 'The Friday Morning Applied Complexity Coffee Group' <[hidden email]>
Subject: [FRIAM] Mawdel Tawk
Dear Wizards,
I commend to your attention a blog post from 538 in which Nate Silver talks with the guy who leads the University of Washington health metrics modeling operation. I understood barely one word in five, but the chief difference between them
seems to be on the degree to which they rely on priors or curve fitting. You folks will, I predict, know what that means. Don’t hesitate to explain it to me.
Nick
Nicholas Thompson
Emeritus Professor of Ethology and Psychology
Clark University
https://wordpress.clarku.edu/nthompson/
| Free forum by Nabble | Edit this page |