The Effects of Large Group Meetings on the Spread of COVID-19: The Case of Trump Rallies by B. Douglas Bernheim, Nina Buchmann, Zach Freitas-Groff, Sebastián Otero :: SSRN
Though it is all but certain that close-spaced largely maskless groups provoked excess infections in the counties in which they occurred compared with other counties with similar infection rate growth, I am not certain that this study is particularly convincing.I noticed a couple of circumstantial & procedural peculiarities.1) Publication timing: when one observes FBI public announcements affecting public cerebration of political issues in the immediate pre election period in 2016 & 2020 one questions motivation. In the same way, one notes that this study published in the pre-2020 election period.2) Placebo data of a time-displaced nature led to data that sometimes showed NEGATIVE correlations with infection events and disease.Results of this kind were arbitrarily set to zero when aggregating data.3) Above all, I asked the question: Can I replicate this Regression? The data source was clear at Johns-Hopkins The locations of the counties to be included & excluded were clear. But with a cavalier nod at the dangers of over-fitting data, we are left only with the impression that the parameters were selected based on what worked!
It would be more interesting to see a study of the kind that can diagnose illness using neural network methods, by clustering analysis.Brian W
On Saturday, November 7, 2020, 06:18:53 AM CST, bernard cleyet <bernard@cleyet.org> wrote:
Thanks to Professor Ann Watkins, stats. maven at Cal. State U. @ Northridge.
On 2020/Nov/02, at 13:29, Watkins, Ann E <ann.watkins@csun.edu> wrote:
bc … didn’t open the link in original post to the working paper.
_______________________________________________
Forum for Physics Educators
Phys-l@mail.phys-l.org https://www.phys-l.org/mailman/listinfo/phys-l