Stop: Before you read any further, take 30 seconds, study the graph above, and identify when Canadian Thanksgiving occurred, based on the huge surge in cases that occurred two weeks afterwards. Answer given at the end of this posting.
After my last post (#915, Still no post-Thanksgiving surge), my wife reminded me that Canada had seen a wave of new cases following Thanksgiving. And so it was and is reasonable to think that we would see such a wave.
Sure, I recall reading that. I bet we all read something to that effect. I didn’t even think to question it. After all, it was on the internet, so it has to be true.
Well, not so fast. If you turn away from the newspaper reporting, and actually start reading what the evidence was, for that “surge” in cases following Canadian Thanksgiving — and in particular, read exactly what the key Canadian epidemiologist said — it’s clear that there was no evidence of a surge.
Let me start with a quote from the epidemiologist who apparently played a key role in determining that this “surge” occurred. And then I’ll take take one minute to explain why that’s no evidence at all. Emphasis mine.
"It’s not that we were flat and all of a sudden Thanksgiving happened and there we see an increase," Laura Rosella, an epidemiologist at the Dalla Lana School of Public Health at the University of Toronto, tells Time. "The reason why we’re fairly confident Thanksgiving did increase cases is that we saw our highest numbers yet in the two weeks following Thanksgiving, which is consistent with the incubation period, when people would show symptoms and get reported."
Source: This was buried in an otherwise over-the-top article in Smithsonian magazine.
As a fellow who did statistical analysis of health care data for his entire career, I cannot even begin to tell you how unforgivably bad that analysis is. They seem to say that because cases were higher after Thanksgiving, that must be due to Thanksgiving. And they didn’t even get the timing right. Those cases wouldn’t be in the data during the two weeks after Thanksgiving, the bulk of them would show up starting around two weeks after Thanksgiving (as explained in my just-prior post.) And, most unforgivable of all, they didn’t even bother to try to account for the pre-existing upward trend in cases.
Words fail me for how bad that analysis is.
Let me just apply the same logic to Virginia. Here, due to an upward trend, we’ve been setting new records now, pretty much every week, for roughly the past two months. You could literally pick any day, in the last 10 weeks, and say, see, cases in the two weeks after that day set a new record. And that would be true. What would NOT be true would be to conclude that something special happened in Virginia, on that day, to cause the subsequent increase.
(And it gets worse. If you read on, in the Smithsonian article cited above, you find this gem: “Nearly two weeks after Thanksgiving, the two most populous provinces in Canada set new records for Covid-19 cases.” As if, what, they didn’t have Thanksgiving in the other provinces and territories?)
Answer: So, just use your eyes, folks. Below, here’s the same graph as at the start of this post, with a line marking Canadian Thanksgiving this year. If you see some big, well-defined post-Thanksgiving surge in cases, starting about two weeks after Canadian Thanksgiving, you have more imagination than I do.
Heck, just to show you how dangerous Thanksgiving is, look at the post-Canadian Thanksgiving surge that occurred in the U.S.A. (below). Yeah, that’s sarcasm. But that’s literally the exact same logic that lies behind the conclusion that Canadian Thanksgiving led to a surge in cases.
So I think I’m going to stop apologizing for not finding any post-Thanksgiving surge in the U.S. data. Yet. The idea that we ought to see one, because Canada saw one, is simply false.
