The internet home for Hoos anywhere and everywhere.

The Corner

Faz d. Hoo

Joined: 8/20/98 Posts: 10555
Likes: 7367


The thing you have to recognize is how little the models actually say


First, they have obvious limitations. The modelers themselves admit this. ICL says "Therefore, we illustrate the impact of likely near-term scenarios for Rt over the next 8 weeks, under assumptions of a relaxation of interventions leading to increased mobility. We note that mobility is acting here as a proxy for the number of potentially infectious contacts. Our mobility scenarios do not account for additional interventions that may be implemented, such as testing and contact tracing, or additional behavioural modifiers, such as increased use of masks. Given these limitations, our scenarios should be considered pessimistic." (see link)

Second, they have levels of uncertainty you can drive a truck through. If you take the midline of their projections, things do indeed look bleak. However, in the case of this model, the confidence intervals straddle an R of 1 quite widely for basically every state. Even the discredited IHME model is well within the 95% confidence interval it established for itself. That interval just happens to be huge.

The reason this announcement attracted my attention in the first place was a statement in an FT report that said "...warned that around half of US states still have Covid-19 reproduction rates above one". When I went to the source, I discovered that the FT was attributing an undue amount of certainty to the study. So there you go. The modelers release a report, then they summarize it, then the media summarizes the summary. The reader is left with an incorrect perception of what the study did and didn't do. Multiply that by M studies and N media outlets and it is no wonder no one believes anyone.

(In response to this post by WahooFla)

Link: ICL overview


Posted: 05/22/2020 at 6:09PM



+2

Insert a Link

Enter the title of the link here:


Enter the full web address of the link here -- include the "http://" part:


Current Thread:
  The model that predicted 2M US deaths? -- 111shamrock 05/22/2020 9:47PM
  I hope those models are wrong. They look grim ** -- hooincarolina 05/22/2020 5:32PM
  I doubt any of these models have had proper V&V. -- The Hook 05/23/2020 5:31PM
  Modeling depends on data, and data depends in part on testing -- NorthernExposure 05/22/2020 5:06PM
  God forbid there be any positive news. ** -- hooincarolina 05/22/2020 7:45PM
  Also true... ** -- Faz d. Hoo 05/22/2020 7:32PM
  140 dollars a test x 300,000 tests a day= -- hooincarolina 05/22/2020 7:12PM
  Lol - we can reproduce more people. wth? -- Hoodlum 05/22/2020 7:49PM
  You can’t continuously print more money -- hooincarolina 05/22/2020 7:44PM
  Yeah, here's one such story -- chicken 05/22/2020 4:31PM
  Then the peers were found wrong. ** -- Hoogle.com 05/22/2020 10:40PM
  Don't shoot me, I'm just the messenger ** -- Faz d. Hoo 05/22/2020 4:59PM
  I know ** -- Toolie92 05/22/2020 5:03PM
  Your story is about the IHME model ** -- Toolie92 05/22/2020 4:34PM

Get the Sabre Edge!

Subscribe to TheSabre.com!
vm311