Bayesian inference and forecast of Covid-19 in Germany by a Max-Planck-Institute

Posted by freemint on Friday May 22, 2020 @05:11PM
from the github.com dept.

https://github.com/Priesemann-Group/covid19_inference_forecast/blob/master/README.md

( 3 comments | 2 points | on HN )

Re:Bayesian inference and forecast of Covid-19 in Germany by... by vital303 on Friday May 22, 2020 @06:36PM (#23277855)

I am puzzled by the paper of this groups [1]. The only evidence they present for "change points in the COVID spreading rate" is slightly better fit of the number of daily reported cases. In this repository for "Scenario assuming three change points with a weekly modulation of reported cases" they use 16 parameters to fit a smooth line (wavy pattern is because of lower case reports around the weekend). An elephant needs 8 [2]!. I also asked for clarification on stats.stackexchange [3].

[1] http://dx.doi.org/10.1126/science.abb9789

[2] http://dx.doi.org/10.1119/1.3254017

[3] https://stats.stackexchange.com/q/467920/286008

[ Parent | on HN ]

Re:Bayesian inference and forecast of Covid-19 in Germany by... by freemint on Friday May 22, 2020 @05:14PM (#23277123)

This repo is maintained the Max-Planck-Institut für Dynamik und Selbstorganisation (dynamics and self-organization). The preprint is available here: https://arxiv.org/abs/2004.01105

[ Parent | on HN ]

Re:Bayesian inference and forecast of Covid-19 in Germany by... by freemint on Friday May 22, 2020 @05:24PM (#23277235)

An interview with the leader of this group is available here: https://youtube.com/watch?v=nI2_AFzE_ek

[ Parent | on HN ]