Odd that they chose to put GDP change on the X and deaths on Y while suggesting a deaths on GDP causation chain. Again their way implicitly suggests that the resilience of the economy has causation on death rates, which is plausible but not the conclusion I think they prefer.
Any way to easily port to an app that gives R and p for the more complete set?
Maserschmidt has this down now, so I don’t want to continue to step on his toes. He can include the model description for his trendline which has all sorts of metrics behind it.
Yeah I am also surprised that I’ve not seen this picked up in the media at all, and that before you the only reaction here was to my “homogenous” (which is consistent with my more common way of saying it, not a typo) over “homogeneous”.
It is big and potentially very good news. But crickets.
A discussion of why it doesn’t make the top of any sources news cycle, or garner much response, would be interesting, but a needless distraction.
You can mouse over the trend line to see R and p, at least from my view of it…let me know if not. I’ve also looked for ways to make it permanently display (like in Excel), but not seeing that.
That looks like a nice study. However, Iceland did a great job with covid so they never had a lot of cases. Their official number of cases is about 0.5% of the population (only 2141 people with 10 deaths). So their official cases are only half what this study is saying (0.9% of the population). Therefore, their case fatality rate is less than double the infection fatality rate (0.5% vs 0.3%). Iceland had widespread testing and contact tracing. I’m sure there are shitloads of serological studies continuing in the US. The data will be interesting.
So R squared is 0.28? I’m no statistician so maybe I’m mistaken, but that doesn’t seem like a strong association. But the p value looks to be pretty good. How does that happen?
Sure, but the main takeaway from the study is not infection fatality rate or anything else you mention - by far the most important result there is the evidence for sustained humoral immune response (at least 4 months, the length of the study) in a high proportion of cases. I see no reason why the smaller total number of cases in the population would affect that. The principal reservation would be whether there are genetic peculiarities in the Icelandic population, but that’s a small probability.
I’m not disputing that at all. In fact, I wasn’t buying in to the scary headlines about waning immunity. Unless, sars-cov-2 was going to be a very unusual virus, memory cells would remain even if antibody levels decline. In fact, plasma cells (the ones that pump out the antibodies) can hang around in the bone marrow for decades. The cases of reinfection are a little concerning but these seem to be rare out of the over 20 million covid cases.
A small(-ish) R squared just means there are many other factors not in the analysis which influence GDP - not a surprise when you include just one predictive factor for something complex like the economy.
I’m not a Tableau person (though maybe I should become one…) but I loaded up those numbers in R and analysed it again with GDP change against the log of per-capita deaths. R squared jumps up to 0.45 and the p-value for the predictor logPerCapitaDeaths is 4.48e-06
27,292,583 total cases
887,554 dead
19,377,267 recovered
In the US:
6,460,250 total cases
193,250 dead
3,725,970 recovered
Yesterday’s numbers for comparison:
Another slice of this same data with the same conclusions but with more commentary. There are two charts in the article. The one in the thumbnail is the second chart. Their conclusion is that they see no evidence of a trade-off between protecting the economy and protecting people’s lives. Mostly, they go in tandem. That goes along with the results of the 1918 pandemic studies as well.
Aspidistra covered the other variables at play, and hinted at the fact that it’s not a perfect linear relationship. If you simply change the trend line from linear to exponential, R squared goes to almost .45, as seen below.
https://public.tableau.com/profile/dmc7541#!/vizhome/GDPandCOVID_15994146814380/Sheet2
I forgot to post the big news of the day: India has surpassed Brazil and is now 2nd for total cases (behind the US).
Friends of mine were planning a big anniversary celebration trip to Hawaii for over a year. Talk about a rollercoaster ride! September 2020 seemed far away. Trip finally canceled.
I can completely agree that there is nothing in the available GDP data to date that supports a (silly) contention that protecting health trades off the economy. Of course those who express concern about severe prolonged mitigation choices causing harms are at least sometimes thinking other than in simplistic binary terms and are usually looking beyond just Q2.
Doing no mitigation in order to protect the economy would be absurd. Contending the polar opposite, that the most severe and prolonged mitigation would lead to the best long term economic result is no less absurd.
To me what jumps out of the graph remains that middling contractions are associated with death rates all over the map. Countries whose economies very easily crumped greatly had large death rates, and countries whose economies were solid enough to stay pretty stable or grew, had low death rates. Fewer died in more resilient economies. More died in more vulnerable ones. The associations really mainly show up only at the GDP change extremes. S shaped fits best. Really can’t assume chicken or egg.
FWIW how many accept China’s data (both deaths and GDP) on face value at this point?
I don’t believe China’s death rate or GDP. Given that, I took them off the chart.
Getting to the point on reactions/lockdowns, I found a lockdown “stringency index” produced by the University of Oxford, which gets updated regularly (4 metrics, 0-100 scale, higher is more stringent). I color-coded the graph to reflect it…I don’t know enough about the index to vouch for it, but the results are interesting.
Indeed.
Would it be easy to graph GDP change v stringency index?
Thank you for your data crunching so far by the way!!
Very nice, thanks.
Here it is, glad to do it…it’s very easy, and I’m learning from the data.
There’s a slope, but the dispersion is huge.
Thanks!
FWIW that to me makes the case more strongly. The association, to the degree it appears to exist, is less driven only by the extreme GDP change cases.