heywood on 29/5/2020 at 15:21
That trend looks suspiciously clean.
Where did the estimated number of infections on lockdown day come from? Nobody has any data to ground such extrapolations. We haven't even sampled enough of the population to make a good guess of current infections.
Where did they get excess deaths? I know that for the US, on May 27 (the date on the graph), excess death counts were only available up to May 9, and the May 2 and May 9 data was incomplete.
How did they come up with a lockdown day for the US when it was the states that imposed lockdowns? They didn't do it all at the same time or to the same degree, and a handful didn't do it at all.
Why is Sweden even on this graph when it didn't shut down?
Quote Posted by Dia
Meanwhile, here in the U.S. we're starting on our second wave of this accursed epidemic. But hey, boy howdy, let's all head to the local pool, beach or bar and make sure our rights aren't being trampled on, right? Selfish fucking morons. *smh*
Most states are still in the middle of the first wave. At least a dozen of them haven't even reached a peak yet. You live in one of those states, right? Wisconsin, where the state Supreme Court struck down the stay at home order a few weeks ago and now we're seeing the expected rise in cases? If so, you have my sincere sympathy. I would not want to live there this year.
Starker on 29/5/2020 at 15:35
Quote Posted by heywood
That trend looks suspiciously clean.
Where did the estimated number of infections on lockdown day come from? Nobody has any data to ground such extrapolations. We haven't even sampled enough of the population to make a good guess of current infections.
Where did they get excess deaths? I know that for the US, on May 27 (the date on the graph), excess death counts were only available up to May 9, and the May 2 and May 9 data was incomplete.
How did they come up with a lockdown day for the US when it was the states that imposed lockdowns? They didn't do it all at the same time or to the same degree, and a handful didn't do it at all.
Why is Sweden even on this graph when it didn't shut down?
You might want to read the article itself and the fine print below the charts. It's explained in more detail there.
Quote:
The data were compiled from national statistical agencies for 19 countries for which sufficient information exists to make robust comparisons. The figures include all of the European countries hit hard by coronavirus. The periods for comparison are from when death rates in individual countries climbed above five-year averages.
The FT has made these comparisons for the first time because the level of deaths in other hard-hit European countries, such as Italy and Spain, has returned close to the seasonal norm.
Quote:
Excess mortality is calculated by counting everyone who has died in a country and subtracting the average number of people who passed away over the same period in the past five years.
It therefore tallies the number of people who died either directly from Covid-19 or indirectly, for example if they were unable or unwilling to seek treatment in hospital, and does not reflect different testing regimes for the virus in different countries.
Quote:
Where a country did not lock down, lockdown-day refers to when transit usage fell below 50% normal levels.
Here's the updated chart, btw, with population numbers taken into account:
(
https://www.ft.com/content/6b4c784e-c259-4ca4-9a82-648ffde71bf0)
Inline Image:
https://www.ft.com/__origami/service/image/v2/images/raw/https%3A%2F%2Fd6c748xw2pzm8.cloudfront.net%2Fprod%2F90f26d70-a0d0-11ea-9bec-2b7f5ee3f0e9-standard.png?fit=scale-down&quality=highest&source=next&width=700
Starker on 29/5/2020 at 15:58
BTW, I often just clip out the most salient points and leave a link to the original source, so I'd encourage anyone to go to the source of the things I post and read them for yourself for the full picture. I've been trying to mark the omissions, so that it would be clear, but in this case, I guess it wasn't really obvious that the footnotes below the graph explained the things on the graph, because they were so small. But anyway, the full article is here: (
https://www.ft.com/content/6b4c784e-c259-4ca4-9a82-648ffde71bf0)
heywood on 29/5/2020 at 18:32
Thanks.
I read the article and unfortunately, it doesn't answer any of my questions. Worse yet, they don't provide any information about their data sources at all.
I'm questioning it because I know for a fact that US excess deaths are compiled weekly, but the reporting lags by about two weeks. Also, the weekly totals are incomplete when first reported and revised up afterward as late reports trickle in. So the last three available weekly totals (May 2, 9, 16) should not be considered reliable. I don't know how other countries count & report excess deaths, but there's probably others for which it lags too. I was hoping the article would say which date the excess death counts were valid for, but it doesn't. It only says "Data updated May 27". Their excess deaths total for the US is less than the official count of Covid-19 deaths as of May 27. The same looks to be true for a few other countries, so if the data is real, it's can't be current through May 27. Without giving a validity date or attributing their sources of data, nobody can fact check this.
For countries that didn't have a lockdown day (Sweden) or didn't have a single lockdown day (US), they don't bother to say what day they used. I Googled around for a bit to find data about Sweden's transit usage during the pandemic, and the only source I found was based on Android phone location data: (
https://www.gstatic.com/covid19/mobility/2020-05-21_SE_Mobility_Report_en.pdf)
But it only goes back to April 9.
Further, there's no explanation of how the number of infections on lockdown day were estimated. That's the statistic I find particularly dubious. Note they are not using the official numbers of confirmed cases on lockdown day, they are trying to estimate the total number of infected, including those who are asymptomatic, or symptomatic but did not get tested. None of the countries on this graph have enough testing capacity to even measure this today, let alone back in March. The best we could do is to perform a controlled antibody test on a broad-based random population sample to estimate what percentage of the population has been infected. But we haven't done that, and even if we had, it wouldn't tell you when people were infected. So you'd have to extrapolate backwards using positive PCR data over time as a guide, and we know you can't compare that country to country (or even region to region within a country) because we've been testing at different rates.
The other thing I find wrong with the whole premise of this comparison is that different countries are in different stages of this first wave. Most of the countries they're using from mainland Europe are on the tail end. But the UK is not there yet, the US is further behind, Sweden further still, and Chile and Peru are still on the front end of the curve.
And I am sorry Starker, I feel like I'm shooting the messenger here. I just thought FT was a somewhat reputable source, but this has me thinking otherwise. Given all the differences between countries' lockdown measures, population density, and how far along the curve they are, I found it rather unbelievable that all the countries would line up so tightly along the regression line, which conveniently supported the authors' criticism of the UK's response. And then oops, we forgot to control for population, and now the correlation mostly falls apart. Seems very suspicious to me. And unable to be fact checked.
Starker on 29/5/2020 at 19:41
Yes, as I said the first time, you should be careful in how you interpret this data, since some countries are still accumulating excess deaths at a very rapid pace. And yes, the numbers will be revised for months and years to come and this is at most a very rough picture based on publicly released and likely preliminary figures. But that's the nature of the beast of trying to assess something still in progress, isn't it? At least until all the data is in and has been revised, which will probably take years and years. Could they be clearer about their sources, sure, but you simply cannot expect the level or rigour from a newspaper article that you would get from a scientific paper. Speaking of which, I think it's also a bit unfair to say that they "forgot to control for population". That data was always there in the article, even before they revised that one chart. In fact, their very first chart in the article reflects this (and always did).
Also, even if you doubt FT's intentions, I don't think there's any doubt that UK has been hit hard by this epidemic, having more cases and deaths than any other country in Europe and being second only to the US worldwide. And I think there's a good case to be made that the slow response and the initial plan of going for herd immunity contributed to that.
Also also, The Guardian reached pretty much the same conclusion (and seems to have links to their sources):
Quote:
(
https://www.theguardian.com/world/ng-interactive/2020/may/29/excess-deaths-uk-has-one-highest-levels-europe)
Patterns in the data show countries that locked down earlier tended to have fewer deaths. Austria, which imposed strict containment measures on 16 March, when there was just one death attributed to Covid-19 in the country, recorded a peak in excess deaths of 14%. By contrast, the Netherlands waited until its excess deaths were already 17% higher than usual before locking down, and at its peak the death toll was 74% above average.
Tony_Tarantula on 30/5/2020 at 02:42
Quote Posted by heywood
Thanks.
And I am sorry Starker, I feel like I'm shooting the messenger here. I just thought FT was a somewhat reputable source, but this has me thinking otherwise. Given all the differences between countries' lockdown measures, population density, and how far along the curve they are, I found it rather unbelievable that all the countries would line up so tightly along the regression line, which conveniently supported the authors' criticism of the UK's response. And then oops, we forgot to control for population, and now the correlation mostly falls apart. Seems very suspicious to me. And unable to be fact checked.
I'm not sure it does line up neatly along the line. The graph exhibits noticeable negative autocorrelation of the residuals but unfortunately they're not giving us quite enough data to run a Durbin-Watson test across the entire sample.
jkcerda on 30/5/2020 at 19:22
virus is gone, millions are out there burning & looting.
Nicker on 30/5/2020 at 20:12
mods
Renzatic on 30/5/2020 at 20:23
On it.
Nicker on 30/5/2020 at 20:35
Thank you.