Page 122 of 258
Re: COVID-19
Posted: Tue May 05, 2020 3:02 am
by Millennie Al
Yes. I think the main thing we can learn from it is that if you put enough mathematics into a model it may distract people from noticing that you have also put in lots of parameters which have been estimated, so the output is highly dubious. I notice that it defines lockdown as "away from work" (L
j(t)) when it normally means a lot more.
It's very much an exercise in showing that, if we could arrange that X is the case, then we would get result Y. Before relying on a result like that I'd like to see evidence that the X can be achieved and what are the likely results of plausible failures to achieve it. We have seen evidence of the effectiveness of lockdown, and contact tracing in other countries. Before adopting a different strategy the alternative strategy should be tested in a small area. After all, if we were talking about treatment for individuals, it would be considered highly unethical to just adopt widespread treatment based on theoretical analysis. Why should it be different on large populations?
Re: COVID-19
Posted: Tue May 05, 2020 5:55 am
by Grumble
Re: COVID-19
Posted: Tue May 05, 2020 7:20 am
by Woodchopper
The early detection and isolation of cases was estimated to have prevented more infections than travel restrictions and contact reductions, but combined NPIs achieved the strongest and most rapid effect. The lifting of travel restrictions since February 17, 2020 does not appear to lead to an increase in cases across China if the social distancing interventions can be maintained, even at a limited level of 25% reduction on average through late April.
https://www.nature.com/articles/s41586- ... _GL_Nature
tl;dr test, test, test
Re: COVID-19
Posted: Tue May 05, 2020 10:21 am
by jimbob
From following the link, and what Twitter was showing nearby
https://twitter.com/jdawsey1/status/125 ... 27232?s=20
Um... That's really not going to happen
Re: COVID-19
Posted: Tue May 05, 2020 10:26 am
by headshot
The headline of that article states 200,000 deaths PER DAY by June 1st. What are they talking about!?
Re: COVID-19
Posted: Tue May 05, 2020 10:32 am
by lpm
3,000 per day, about 200,000 total, is what they meant to say.
But the White House is claiming there's a cubic model that shows US cases will disappear at the end of May, early June, then become negative.
Re: COVID-19
Posted: Tue May 05, 2020 10:34 am
by headshot
Their f.cking HEADLINE:
Re: COVID-19
Posted: Tue May 05, 2020 10:51 am
by shpalman
lpm wrote: Tue May 05, 2020 10:32 am
3,000 per day, about 200,000 total, is what they meant to say.
But the White House is claiming there's a cubic model that shows US cases will disappear at the end of May, early June, then become negative.
Does that mean they just fit a third-order polynomial in Excel?
Re: COVID-19
Posted: Tue May 05, 2020 10:55 am
by Woodchopper
lpm wrote: Tue May 05, 2020 10:32 am
3,000 per day, about 200,000 total, is what they meant to say.
But the White House is claiming there's a cubic model that shows US cases will disappear at the end of May, early June, then become negative.
Nope, its 200 000 cases per day. That's what exponential looks like.
From the 'report' which is actually a PowerPoint. Linked to from the Washington Post link above.
- Projection.jpg (212.85 KiB) Viewed 4274 times
Re: COVID-19
Posted: Tue May 05, 2020 10:59 am
by RobS
Does that mean they just fit a third-order polynomial in Excel?
It would appear so:
https://twitter.com/NateSilver538/statu ... 18470?s=20
https://twitter.com/NateSilver538/statu ... 87682?s=20
- EXN0eznWoAAITlP.jpeg (28.63 KiB) Viewed 4272 times
Re: COVID-19
Posted: Tue May 05, 2020 11:01 am
by headshot
Oh, WAIT A MINUTE. The headline I screenshot said 200,000 deaths per day when I read it earlier. It must have been corrected just before I screenshot it...
Re: COVID-19
Posted: Tue May 05, 2020 11:26 am
by shpalman
headshot wrote: Tue May 05, 2020 11:01 am
Oh, WAIT A MINUTE. The headline I screenshot said 200,000 deaths per day when I read it earlier. It must have been corrected just before I screenshot it...
There seems to have been confusion between 200,000 deaths in total and 200,000 cases per day.
Re: COVID-19
Posted: Tue May 05, 2020 11:41 am
by Holylol
Well, then I have a model to sell that perfectly fits the data for the last 30 days... it's a 29th order polynomial.
Re: COVID-19
Posted: Tue May 05, 2020 11:42 am
by Woodchopper
shpalman wrote: Tue May 05, 2020 11:26 am
headshot wrote: Tue May 05, 2020 11:01 am
Oh, WAIT A MINUTE. The headline I screenshot said 200,000 deaths per day when I read it earlier. It must have been corrected just before I screenshot it...
There seems to have been confusion between 200,000 deaths in total and 200,000 cases per day.
Yes, at about 2000 per day we'll be around 200 000 deaths by the end of June. The problem for the US is that deaths appear to have plateaued rather than started to decline.
Re: COVID-19
Posted: Tue May 05, 2020 11:44 am
by lpm
Holylol wrote: Tue May 05, 2020 11:41 am
Well, then I have a model to sell that perfectly fits the data for the last 30 days... it's a 29th order polynomial.
Does it agree with the White House version and show negative deaths in June? If not, you've done something wrong.
Re: COVID-19
Posted: Tue May 05, 2020 12:55 pm
by Holylol
I can't be arsed to try as I don't have the data points. It will either go to +infinity or -infinity with time...
Re: COVID-19
Posted: Tue May 05, 2020 12:59 pm
by bmforre
lpm wrote: Tue May 05, 2020 11:44 am slightly adjusted:
Does it agree with the
Evangelical White House version and show
negative deaths in June?
But surely by then Easter is well past for this year?
Re: COVID-19
Posted: Tue May 05, 2020 2:45 pm
by shpalman
Holylol wrote: Tue May 05, 2020 11:41 am
Well, then I have a model to sell that perfectly fits the data for the last 30 days... it's a 29th order polynomial.
- HoC.png (21.64 KiB) Viewed 4161 times
https://xkcd.com/2048/ of course
Re: COVID-19
Posted: Tue May 05, 2020 2:48 pm
by PeteB
Institute and Faculty of Actuaries (IFoA)
Continuous Mortality Investigation estimate 55,000 excess deaths for UK so far
Re: COVID-19
Posted: Tue May 05, 2020 2:48 pm
by shpalman
Woodchopper wrote: Tue May 05, 2020 10:55 am
lpm wrote: Tue May 05, 2020 10:32 am
3,000 per day, about 200,000 total, is what they meant to say.
But the White House is claiming there's a cubic model that shows US cases will disappear at the end of May, early June, then become negative.
Nope, its 200 000 cases per day. That's what exponential looks like.
From the 'report' which is actually a PowerPoint. Linked to from the Washington Post link above.
Projection.jpg
Great news everyone! Those deaths so far didn't actually happen!
Or am I reading it wrong? If there's a link in the WashPo to the actual thing, can someone post the link here please?
Otherwise I would assume those registered deaths did happen (they're of course an underestimate but we'll worry about that later) and scale the actual number of cases up to take into account that there hasn't been enough testing, rather than assuming the number of cases is correct and ignoring the scale of the death toll.
Re: COVID-19
Posted: Tue May 05, 2020 2:51 pm
by Woodchopper
shpalman wrote: Tue May 05, 2020 2:48 pm
Woodchopper wrote: Tue May 05, 2020 10:55 am
lpm wrote: Tue May 05, 2020 10:32 am
3,000 per day, about 200,000 total, is what they meant to say.
But the White House is claiming there's a cubic model that shows US cases will disappear at the end of May, early June, then become negative.
Nope, its 200 000 cases per day. That's what exponential looks like.
From the 'report' which is actually a PowerPoint. Linked to from the Washington Post link above.
Projection.jpg
Great news everyone! Those deaths so far didn't actually happen!
Or am I reading it wrong? If there's a link in the WashPo to the actual thing, can someone post the link here please?
Otherwise I would assume those registered deaths did happen (they're of course an underestimate but we'll worry about that later) and scale the actual number of cases up to take into account that there hasn't been enough testing, rather than assuming the number of cases is correct and ignoring the scale of the death toll.
Here's a link:
https://context-cdn.washingtonpost.com/ ... d2.#page=1
Re: COVID-19
Posted: Tue May 05, 2020 3:11 pm
by Woodchopper
On UK mortality, still rising in care homes, and still rising in some regions.
https://twitter.com/henryjameslau/statu ... 7287365633
Not one curve, but lots of different ones.
Re: COVID-19
Posted: Tue May 05, 2020 6:09 pm
by shpalman
Possibly they meant "cubic splines" so that's OK then.
https://twitter.com/CT_Bergstrom/status ... 60615?s=19
Re: COVID-19
Posted: Tue May 05, 2020 6:10 pm
by shpalman
Re: COVID-19
Posted: Tue May 05, 2020 6:15 pm
by dyqik
Of course, today's death toll is already pushing the most pessimistic of those models, the IHME (5/4) model, presumably released yesterday.