Are The Lockdowns Working/How Effective Are They?

[jim360]

As is usually the case, the answers are "possibly" and "fairly, maybe?". But this is also an interesting video to see how statistical analysis can be used to even try to answer the question.

[fiction-factory]

Watching it now. Up to minute 7 . Will continue. Just didn't want you to think no-one was interested

[fiction-factory]

Thanks jim. I enjoyed that. I was a bit worried by the Chinese sounding accent at first and was expecting some propaganda but it seemed a fair analysis and I liked the presenter's approach. It has limitations but I suspect or at least hope that the real stats wizards are doing something similar and refining the models. I will have tow watch again as I'm not sure why the green lines were falling towards 1 even before lockdown, but the presentation tells me we need to be really clear about getting the rate well below 1 and that lockdowns must be kept in place until we are sure it's safe to loosen them. It's a shame they charts didn't cover Spain, UK, Germany. I'm also not sure why USA went to 1st April only- I thought it could have gone on another week.
I would be really interested to see this applied to UK data.
I still have to think though about what the data on infections tells us because that is dependent on the testing regimes.

Thanks. Will watch again later if I get chance.

I hope PP sees it

[jim360]

American-born Chinese person, I believe. No propaganda -- I suppose in accepting the Hubei/Wuhan figures there's something there but there's not much choice but to start from data you *do* have, and that it is reasonably consistent with the statistical model suggests that it's worth taking at least somewhat seriously.

Clearly not meant to be the definitive analysis, of course.

[BilSuth]

I understand that there are two elements to lockdown effect. If you stay at home like a hermit, the infection risk is 0. Even I get that.
Clearly the lockdown reduces interaction and the effect of that has been to flatten the graph but at an unacceptably high level.
The second element is how many infected people are out there to avoid interacting with. FF (always valued) in his second contribution refers to the R figure, how many others on average the virus is transmitted to by an infected person. Pre-lockdown this was said to be 2.4 (or even up to 4) and getting it below 1, as claimed now, was a big objective. I takes longer to manifest itself.
X x Y becomes less scary when X (human movement) is reduced and Y (those infected) becomes 1 in 500 instead of 1 in 50.

[fiction-factory]

I wonder what would happen to the figures if we had as near as possible to a full lockdown for maybe a week or two, including no shops at all except pharmacies, no travel without a permit, no deliveries other than maybe one food parcel per house. Would be painful for many or most but I'm wondering whether it could really help slash the number of carriers and any identified could be put in quarantine.

[BilSuth]

Maybe we could, FF. Next question is how could we keep it at 0 when we failed when infections started and were still low? Same applies to every other country who let the virus take hold. We would need mandatory testing at airports and ports for a start.
As for getting to 0, even flagship South Korea have a handful of deaths every week. Their policies have included massive testing plus contact tracing. The down-side has been victimisation and trolling of suspected infected/carriers. It ain't easy.

[jim360]

Watching the video inspired me to finally model the progress of the disease properly, or at least try to. Managed to create a reasonably good fit to the NHS England daily deaths data, although I'm going to need to do a lot of tweaking to improve the fit.

It's an unfortunate and macabre side-effect that I'm learning much more about statistics now than I have in a long time.

[Zacs-Master]

We look forward to your results, Jim.

[fiction-factory]

It would be great if you could share it jim.
Hi Bill- I'm not expecting to get to 0 but it must be possible to get a big reduction by a short spell of hermit living for everyone, couple some sort of test, track and trace system plus enforced isolation for detected cases.

Yes the lockdown is working pretty well but I've noticed things slipping in the last week - e.g. I've seen a lot more family visits/overnight stays at neighbours , teenagers gathering in groups again and people going shopping twice a day for something to do, full day outings to beaches/picnic spots - and there's a risk things may not fall as much as hoped for unless we have a short strong tightening to shift things downwards more quickly.

[jim360]

It should go without saying, of course, that my results are going to be nowhere near as sophisticated as many other analyses out there. Only using two parameters for my fit at the moment and I think that's probably all I'd be capable of without a crash course in modern advanced epidemiology.

Also I'm still tweaking, and I'll want to see another week or so of data to see how good the fit is after the supposed peak in early April before I "announce" anything, for all that people here will care.

Pl

[Zacs-Master]

So why bother?

[jim360]

Because I'm learning something while I'm doing it. Next question?

[Neveracrossword]

Learning for learning's sake is surely admirable?

[Zacs-Master]

What’s the meaning of life? And DONT say 42!

[jim360]

"quarante-deux".

:P

[fiction-factory]

If that was to me Zacs, I'm just thinking out loud whether a final push with a 1-2 week more intense lockdown period (most of us living as hermits, no supermarkets/ journeys without a permit)- coupled with a promise of a planned phased relaxation to start immediately afterwards- we get would jolt the curve downwards sufficiently to allow the relaxation to be done safely

[fiction-factory]

Okay Zacs- my typing is too slow. It seems your "why bother" comment was to jim not me

[Zacs-Master]

Just wait till the weather turns bad, ff.

Jim. Cop out! ;-)

[Peter Pedant]

hi boys - just got to this
er was reading - no really - aeschylus - Persians - eye witness account of the battle of salamis 478 BC

good idea - getting the death ratio or case ration and then projecting backwards
I thought just do it with a fraction - and was pretty impressed by his analysis of getting a normal distribution for each daily result
wow !
I can think of it but cant do it
BUT

Jim - if you are gonna do this - please please watch Zhukovs viddie ( he lectures to the NHS providers) and or the Oxford viddies - Both on You tube

Cary Huang above doesnt mention the SIR model and then I thought - that is because he is not aware of it .....
and his model would be BETTER if he took into account what has gone before.

that is there is a certain amount of Cary H rediscovering the wheel

so he jumps from 1 + 1/7 to 1.14 to Ro of 1,14 - which I think is terribly low - and altho he generates valid constants for his own model, they dont correspond to the constants in the other models - - - which makes comparison difficult

and he reconstructs data given in the New England Journal paper which he just could have used ( 5 d incubation with SD of 1) - so he spends around 5 m estimating data measured in the journals

and
// as I'm not sure why the green lines were falling towards 1 even before lockdown,//

yes I noticed that as well - and I thought it might be an effect of the SIR model - you run out of people to infect,
and not an effect of lock down
Cary H didnt mention it and that was why I thought he wasnt conversant with the SIR model
( because even if he thought it wasnt, he should explain why he thought it wasnt. but then he says - oh but he is just a numba cruncher...)

but blimey
very well done
I cant do anything like this....

yeah you see the difficulty about having a decreasing rate before lockdown is ....... if the rate isnt tweaked by lockdown ( angle hump or corner ) - then you can say that lock down hasnt had an effect cant you?

oo-er Mrs ! lucky this is locked away in Science
and no one understands what and why I say

so it was a bit like the Chartists view of the Stock Exchange variations ( you dont look at any of the company data - you just look at the way the numbers(prices) vary - no really they do this - chartists I mean. - shoulders and dead cat bounces and so on)

anyway well done Cary and well done Jim

[douglas9401]

The Persians fought with sausages? Come on now!