This can be a reprint of a post from 2020 entitled “The Worst Statistical Analysis I Have Seen This Year”, motivated by Mr. Bruce Hall‘s urging that we take a look at a Judith Curry link.
(And I’ve seen plenty of horrible evaluation) [Update 8/14/2020: the author has taken down the post, but here is an archived 8/13/2020 version of the webpage]
Reader Bruce Hall recommends hyperlinks to this article which asserts that 2020 just isn’t anomalous when it comes to deaths. In truth, it’s twentieth out the final 21 years!
Now Ms. Colleen Huber, NMD*** involves this conclusion thusly:
As of this writing, 32 weeks have elapsed in 2020. Nevertheless, for every earlier yr, 52 weeks have already elapsed. How then can we examine deaths from all causes in 2020 to earlier years?
I divided the whole variety of deaths for annually by the variety of weeks. That’s 52 weeks for all years, apart from 2020, wherein 32 weeks have elapsed as of this previous Saturday, August 8, 2020, which is essentially the most lately up to date week within the CDC knowledge cited. This offers us the common variety of deaths per week for every of these years, and permits a significant comparability between 2020 and prior years.
She then generates the next desk:
She concludes:
Plainly there is no such thing as a pandemic in 2020 of COVID-19 or of anything, at the least not in the USA.
It’s nice that Ms. Huber tells us there are 52 weeks in a yr. She divides 2020 knowledge by the 32 weeks which have elapsed and have been recorded by CDC (although current weeks are very incomplete when it comes to reporting).
This could be a wise strategy — calculating a per/week fatality charge — if there have been no seasonality within the knowledge. Nevertheless, deaths are seasonal within the US, as can simply discerned within the CDC knowledge she was analyzing.
Determine 1: CDC knowledge accessed 7 August 2020.
As we enter the latter a part of the yr, deaths sometimes rise (with flu, and many others.). Therefore, utilizing 32 weeks for 2020, and all 52 weeks for earlier years, will sometimes yield a nonsensical comparability. (There’s a customary strategy, utilized in many economics releases — year-to-date counts. I.e., Ms. Huber might’ve in contrast deaths within the first 32 weeks of every of the previous 20 years towards these within the first 32 weeks of this yr.)
As soon as once more, essentially the most embarrasingly silly knowledge evaluation I’ve seen this yr (possibly this decade, though the competitors is hard).
My investigation utilizing CDC estimates of anticipated deaths, here.
*** “NMD” means “naturapathic medical physician”
Addendum 9/26: I’m ever grateful to Mr. Corridor for offering this instance of egregiously unhealthy reasoning. I cite it every time I educate PA819 (ungated 2020 website). I’m additionally grateful to Craig Eyermann (aka Ironman at PoliticalCalculations), Steven Kopits, and notably CoRev for many other examples.