Read on Twitter
I& #39;m appreciative of the hard work and courage of teams who have put out public #COVID19 forecasts. This outbreak has been really hard to predict from the outset. There is so much that we don& #39;t know about how all of this will unfold in coming weeks and months. 1/10
I& #39;ve done some spot-checking of forecast accuracy of a few of the US models, and am planning a more thorough analysis of models that have been put out by @IHME_UW @Columbia @LosAlamosNatLab @alexvespi @MRC_Outbreak and others. 2/
On the whole, short-term accuracy over the limited sample size of the last few weeks seems to be not all that bad, but there are pretty major differences in the long-term outlook of these models. I& #39;m going to run just a quick back-of-the-envelope comparison of two models. 3/
Here is a model from Dave Osthus and @sdelvall from @LosAlamosNatLab. Dave has created some the most accurate forecast models for seasonal influenza in the US over the last few years. This model predicts between 18K-176K #COVID19 deaths in NY by 5/20.
https://covid-19.bsvgateway.org/ ">https://covid-19.bsvgateway.org/">... 4/
https://covid-19.bsvgateway.org/ ">https://covid-19.bsvgateway.org/">... 4/
Here is another model from @IHME_UW. This model has been getting a ton of press. To my knowledge, this group has not forecasted outbreaks in recent years. Their latest update predicts that NY will have no more COVID deaths after 5/6, with a total between 9,365 and 21,826. 5/
The IHME model projections are conditional on "full social distancing" continuing through May 2020, which at this point seems like a reasonable assumption, at least through early May. 6/
These two #COVID19 models, both from very well-respected academic groups, are complete odds with one another. Their prediction intervals barely overlap! It highlights for me how little we really know about what will happen in the coming weeks. 7/
What we do know is that social distancing appears to be working. But models about the future are still far from certain. NOAA can barely predict the weather 7 days in advance, so don& #39;t expect the experts to predict #COVID19 death tolls 6 weeks ahead! 8/ https://twitter.com/cmyeaton/status/1248003777325600770">https://twitter.com/cmyeaton/...
It seems that the press has been eager to push the narrative of "we are near the peak!" and "the end is in sight" but given the strong uncertainty about the future and lack of clear consensus among modelers, I think these messages are premature. 9/
https://imgs.xkcd.com/comics/scientific_briefing.png">https://imgs.xkcd.com/comics/sc...
https://imgs.xkcd.com/comics/scientific_briefing.png">https://imgs.xkcd.com/comics/sc...
There are bright spots amidst the crush of tragic data right now, but I think a key storyline right now from mass media needs to be that the future of #COVID19 in the US and globally is very uncertain.
10/10
10/10
