The data in this paper may be systematically messed up. It mostly depends on categorizing counties and countries in various ways versus vaccination rate. Yet the examples the paper gives explicitly as above 90% vaccinated seem very wrong: Chattahoochee (GA), actual 15.2%, McKinley Co (NM), actual 60%; Arecibo (PR) actual 20%. (My data from data.news-leader.com/covid-19-vaccine-tracker). So it might be worth checking if the paper’s wrong attribution of vaccination rate to counties applies across the board. At the very least it suggests very poor quality control or review, since it is surely surprising that these counties would have such high vaccination rates. (Along with Figure 2’s obvious decrease in cases with higher-vaccination rates, strangely at odds with the written narrative.)
His analysis does not take into account any temporal dynamics.
In fact that causes him to possibly confuse cause and effect.
For example, here in Israel, the rise in infections has led to a quick deployment of the booster shot for a very large population.
But this only takes effect after the booster effect starts in the individual, and after enough people are vaccinated in the community. In the meanwhile, the numbers continue rising - so you have a false correlation "Many vaccinated and many sick".
And he picked as a starting point September 3rd, which is the apex of this trend in this particular country. Had he looked a month later, the results would be very different.
Assaf's point about temporal dynamics seems very on-point to me. In the paper's section on Methodology, it says the study compared the total reported cases from one week, to the total cases from the preceding week, and in the example they chose to mention, "Los Angeles county in California had 18,171 cases in the last 7 days (August 26 to September 1) and 31,616 cases in the previous 7 days (August 19–25)". Clearly those two data points, showing a decrease from 31 thousand to 18 thousand in one week (!!), is not overall representative of what was going on in that county on a regular basis that can be meaningfully compared to the vaccination rate to draw conclusions!
The data in this paper may be systematically messed up. It mostly depends on categorizing counties and countries in various ways versus vaccination rate. Yet the examples the paper gives explicitly as above 90% vaccinated seem very wrong: Chattahoochee (GA), actual 15.2%, McKinley Co (NM), actual 60%; Arecibo (PR) actual 20%. (My data from data.news-leader.com/covid-19-vaccine-tracker). So it might be worth checking if the paper’s wrong attribution of vaccination rate to counties applies across the board. At the very least it suggests very poor quality control or review, since it is surely surprising that these counties would have such high vaccination rates. (Along with Figure 2’s obvious decrease in cases with higher-vaccination rates, strangely at odds with the written narrative.)
It's even worse than that.
His analysis does not take into account any temporal dynamics.
In fact that causes him to possibly confuse cause and effect.
For example, here in Israel, the rise in infections has led to a quick deployment of the booster shot for a very large population.
But this only takes effect after the booster effect starts in the individual, and after enough people are vaccinated in the community. In the meanwhile, the numbers continue rising - so you have a false correlation "Many vaccinated and many sick".
And he picked as a starting point September 3rd, which is the apex of this trend in this particular country. Had he looked a month later, the results would be very different.
"the median estimate for cases per 100,000 clearly decreases as addressing rate increases" -- did you mean vaccination rate?
Thank you for the catch!
Assaf's point about temporal dynamics seems very on-point to me. In the paper's section on Methodology, it says the study compared the total reported cases from one week, to the total cases from the preceding week, and in the example they chose to mention, "Los Angeles county in California had 18,171 cases in the last 7 days (August 26 to September 1) and 31,616 cases in the previous 7 days (August 19–25)". Clearly those two data points, showing a decrease from 31 thousand to 18 thousand in one week (!!), is not overall representative of what was going on in that county on a regular basis that can be meaningfully compared to the vaccination rate to draw conclusions!
Germany and UK are not included among the 68 countries selected