The two cool things about this Richard Epstein dude are:
1. He has no capacity for embarrassment.
2. The Hoover Institution has has no capacity for embarrassment in continuing to feature his writings.
When it comes to covid, Epstein is famous for predicting that the number of U.S. covid deaths would reach . . . 500. Which he then upped to 5000. As Rex Douglass pointed out, the problem wasn’t so much that he was off by three orders of magnitude—such things happen, and the number of fatalities from a new pandemic is just flat-out difficult to forecast—but rather that Epstein didn’t even seem to try to learn from his mistakes. Anyway, that was a couple of years ago, but, tenured professor that he is, this didn’t stop Epstein from continuing to pontificate, never reflecting up on the existential question of why anyone should take what he says seriously, except perhaps as implicit advice to do the exact opposite of whatever he recommends.
Here he is earlier this month at his home base, the Hoover Institution, with “The Vaccine Mandate: A Case For Caution,” where he releases fatuous pronouncements such as:
Clearly, if the vaccines were undeniably as safe and effective as the government claims, there would be no need for any mandate at all.
“Clearly,” huh? What an idiot. Jeez. Doesn’t he remember taking his children to the doctor when they were little? There are tons of vaccines that are mandated. I mean, sure, maybe the measles vaccine etc. really is dangerous or ineffective, I’m no expert on the topic—but it’s not “clearly” so!
Extra points for Epstein citing a study in the Epoch Times, (described by Wikipedia as “a far-right international multi-language newspaper and media company affiliated with the Falun Gong new religious movement”), claiming there is now “direct evidence” that the omicron variant “infects unvaccinated and vaccinated people similarly.”
According to Epstein, “That evidence cuts against the claim that the federal vaccine mandate will work well . . .” OK, sure, let’s take any old crap on the internet and use it to strike down a law—that’s cool! Not. Of course, even Pravda will run straight news stories from time to time, but it’s not a great sign if this is the best that he can do for a source.
But . . . I’m confused! According to Epstein, the very existence of the mandate implies that the vaccines are either unsafe or ineffective. So why is he looking for evidence at all?
Also, “That position is also advanced in a letter from the law firm of Siri Glimstad.” That sounds cool. We have a position supported by a propaganda sheet and a law firm. What could possibly go wrong???
Good job, Hoover Institution!
Epstein continues:
The CDC and OSHA might be more credible if they addressed these claims with some thoroughness, but they have consistently declined to respond to them.
Not responding to articles in the Epoch Times or a letter from some law firm??? How shocking.
Again, though, I’m puzzled. What could the CDC and OSHA possibly say that would circumvent Epstein’s ironclad inference: “Clearly, if the vaccines were undeniably as safe and effective as the government claims, there would be no need for any mandate at all.”
Epstein continues:
The administrative law points are elegant, but in modern times it is well also to recall the ancient Hippocratic Oath: first, do no harm.
Good point, dude. After all, Covid is only forecast to kill 500 people in this country. That’s less than the number of people who drown each year in swimming pools! So why can’t we all just chill out.
You can click through to read the whole thing. I skipped some juicy bits like the link to antivax hero Alex Berenson. All-in, baby, all-in.
Why bother?
Why should we care about the Hoover Institution? There are lots of vaccine deniers and vaccine denier sympathizers with a lot more influence than some retired law professor. Indeed, by linking to Epstein’s column I’m just giving him a larger audience.
I think the reason this bothers me—beyond the usual “someone is wrong on the internet” thing—is the nexus between academia, policy, and scientific and statistical evidence.
Even if there were no such thing as Stanford University, Richard Epstein would be influential. During his career, he’s crafted arguments that have been persuasive to government officials—administrators, legislators, and judges. That’s what being an influential legal scholar is all about. And once you have influence in one domain, it makes sense to want to apply it elsewhere. Fair enough. The bad part is when, instead of just saying, “As an eminent legal scholar, my take is that it’s dangerous for the government to regulate X, Y, and Z,” he starts bullshitting about scientific and statistical evidence. I mean, sure, he should be allowed to do this—we still have freedom of speech in this country!—but he shouldn’t want to make a fool of himself in that way. Policy isn’t a debate club where you win by throwing a “spread” of poop at the wall on the theory that some of it will stick—which reminds me, I highly recommend the novels of Ben Lerner; right now I’m in the middle of The Topeka School—it’s real life. To start with a position, and then cruise the internet looking for disinformation sites that will provide pseudo-evidence to support it . . . that’s not good. I again refer you to Rex Douglass’s article, “How to be Curious Instead of Contrarian About COVID-19: Eight Data Science Lessons From Coronavirus Perspective.” But . . . Epstein’s circulation of reheated propaganda is sponsored by the Hoover Institution, which is part of Stanford University, which is supposed to be an institution of learning . . .
Yeah, that bothers me. Even though we can also be bothered about various influential institutions (academia, corporate boards, police unions, etc.) being, in various ways, unrepresentative of the general population. Some big percentage of Americans believe in ghosts, too, but I think that’s foolish. Not consequential, though, I guess. Anyway, these Hoover guys push my buttons—it’s some combination of the connections to political science, statistics, and academia in general. Even though I don’t really have any answers here: on one hand, tenured professors with a political ax to grind will publish somewhere—if it’s not at the Hoover Institution, it might be on twitter, and twitter probably gets more readers anyway—; on the other, given that Epstein’s been doing his shtick for a long time, it’s not like he’d read a post like this and say, “Hey! I can do better.” He sees himself as a scrapper, and I suppose that’s part of the value system of some lawyers, that it’s all about winning, not about learning (again, I recommend that Rex Douglass post).
So, just to summarize this last bit: I write this post out of irritation, not with any hope of changing anyone’s mind, indeed I suspect that very few people will have read Epstein’s article in the first place, but rather as a way of further exploring my understanding of some of the uglier corners of academia, or I should say the intersection of some of the uglier corners of academia, policy, and the news media.
At times, it can be instructive to look at pure stupidity, not because it’s of interest for its own sake, but because its publication reveals something about the system that sustains it. And, as always, we laugh at these horrible things because that’s less painful than crying.
Sometimes the chickens get out of the coop to cannibalize more widely.
Science Is Like A Chicken Coop. Richard McElreath https://www.youtube.com/watch?v=d8LqFO1dk-w&t=1510s
Keith:
Yeah, it’s the social/institutional context—the Hoover Institution—that makes the story notable. Without the Hoover platform, Epstein’s just one more cranky professor sharing nonsensical rants on twitter.
I’m glad that more people are taking note of the failings of the Hoover Institute. The Institute has always been about making sure that Stanford has close ties to powerful political figures and donors in the American conservative movement. At one point the Hoover Institute should have just been called the Stanford Office of Ronald Reagan’s Presidential campaign because they were so integrated into the campaign and fundraising.
Hoover Institute fellows like to attach themselves to the prestige of the Stanford name when it suits them, yet most (all?) of them produce work that is too low quality for them to actually be hired at Stanford. They’re hired explicitly to further a political viewpoint and to raise money from wealthy conservatives despite their lack of academic boba fides, yet these are the same people that will complain about diversity hires on college campuses. Ironic.
Stanford has always quietly tolerated Hoover because it’s useful to be on good terms with conservative political elites. But the Trump administration brought out the worst of Hoover and I don’t think the ugly is going back into hiding anytime soon.
Jeffrey:
It’s not correct to say that possibly all of the Hoover fellows “produce work that is too low quality for them to actually be hired at Stanford.” A bunch of legit Stanford poli sci professors are Hoover fellows as well.
Just in case Andrew’s critique of Epstein’s track record was too subtle, this is from
https://en.wikipedia.org/wiki/Richard_Epstein
————————————————————————-
In March and April 2020, Epstein wrote several essays published by the Hoover Institution giving a contrarian account of the ongoing coronavirus outbreak, and warning against extensive containment and mitigation measures by the US government, which he identified as an “overreaction”.[19] In a piece published on March 16, he argued that the term pandemic is not one to be used lightly and that the virus should be allowed to run its course, predicting there would ultimately be 500 US deaths. In early June the US death total passed 100,000 persons.[20] On March 24, when the fatality number had already passed 500, Epstein added a “Correction & Addendum”, in which he raised his forecast to 5,000 deaths,[21][22][23] without changing the underlying model that had led him to his first estimate.[24] On April 6, when the death toll had already far surpassed his earlier predictions, he again revised that figure, with the “Correction & Addendum” section now suddenly declaring (under the wrong datestamp “March 24, 2020”) that the “original erroneous estimate of 5,000 dead in the US [was] a number 10 times smaller than [he had] intended to state”, implying that both 500 and 5,000 in the earlier article versions had been misspellings of 50,000.[25] After several news reports about Epstein’s ever-increasing estimates, on April 21 an editor’s note appeared on the website, which explained the latest changes as an “editing error” and clarified that the author’s original prediction had been 500 deaths.[26] In December 2020, when the death toll from COVID-19 in the United States stood at over 333,000, Politico named Epstein’s predictions among “the most audacious, confident and spectacularly incorrect prognostications about the year”.
——————————————————————-
I agree with everything that you said. Stupidity regarding COVID-19 and vaccines is on display daily in many places. What annoys me is that they apparently can’t or won’t see that they are treating the vaccines for this virus as somehow categorically different from other vaccines for which we have had longstanding mandates. One suspects that the opposition is driven solely by the imperative to oppose anything favored by the Democrats. But instead of openly acknowledging this as the source of their opposition, they scour the internet for whacky, pseudoscientific arguments against the vaccines. This just adds to the obvious stupidity of staking your political position on opposition to vaccines against a virus whose epidemiological and evolutionary dynamics you don’t understand.
Unfortunately, sars-2 *is* categorically different from measles/polio/etc. It primarily replicates in the mucosa rather than other tissues/blood (viremia is only found in like 25% of severe covid infections). For this type of virus, we never see lasting herd immunity even after infection. And IM vaccines do not even give you that temporary mucosal immunity.
https://pubmed.ncbi.nlm.nih.gov/33901246/
So you appear to be annoyed by reality. Misinformation has been a big problem regarding covid, especially from politicans and the media. The pharma companies know better than to directly spread obvious bs like this.
Here are the first two paragraphs from the reference:
“It is often messaged that herd immunity to Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV2 (CoV-2)), the causative agent of Coronavirus Disease 2019 (COVID-19), will protect nonvaccinated individuals from infection. Herd immunity refers to the concept that when a sufficient fraction of individuals in a population develop immunity from infection or vaccination, viral transmission is reduced to a near negligible level. However, seasonal CoVs, which cause approximately 20% of common colds, remain endemic, even though demonstrating only limited antigenic evolution in epitopes targeted by neutralizing antibodies [1,2]. Although there are only 4 circulating seasonal CoVs, infections frequently recur, even yearly, likely related to waning antibody levels [3]. Human challenge studies established that seasonal CoV reinfection with the identical strain can occur within a year after initial exposure, though typically with reduced shedding and milder symptoms [4]. Reinfection also appears to occur following mild COVID-19 cases, where the serum neutralizing antibody half-life is only approximately 5 weeks [5].
“If immunity to SARS-CoV-2 and seasonal CoVs are similar, COVID-19 herd immunity is a pipe dream, even more so given the relatively rapid selection of mutants with amino acid substitutions in the spike protein that reduce the efficiency of serum antibody neutralization [6]. Absent effective herd immunity, over the next few years, individuals can choose whether their first exposure to SARS-CoV-2 immunogens occurs via vaccination or infection. With the age-related increase in COVID 19 severity, it is critical that individuals be vaccinated sooner rather than later.”
“In a study carried out in Israel, the viral load present in the nasal mucosa of a sample of healthcare workers was evaluated on a weekly basis. The conclusion of the study showed that in vaccinated and COVID-19-positive persons, the viral load was 2–4 times lower than in unvaccinated persons (Callaway 2021). Another study, evaluating the amount of viral RNA present in approximately 16,000 nasal swabs showed that the viral load of SARS-CoV-2 in COVID-19-positive and vaccinated subjects is 1.6–20 times lower than the viral load present in infected and unvaccinated subjects (Levine-Tiefenbrun et al. 2021). Another study performed in the United States was conducted on a sample of 3950 health care workers examined between December 14, 2020 and March 13, 2021. The results showed that vaccines had an efficacy in preventing infection of 90% 14 days after the second dose and 80% 14 days after the first dose (Krause et al. 2021).”
https://www.ncbi.nlm.nih.gov/labs/pmc/articles/PMC8287551/
I’m sure the retort will be that lower mucosal viral load won’t reduce transmission, or that the studies are “silly”.
During May–November 2021, case and hospitalization rates were highest among persons who were unvaccinated without a previous diagnosis. Before Delta became the predominant variant in June, case rates were higher among persons who survived a previous infection than persons who were vaccinated alone. By early October, persons who survived a previous infection had lower case rates than persons who were vaccinated alone.
https://www.cdc.gov/mmwr/volumes/71/wr/mm7104e1.htm
The relative effectiveness against symptomatic disease 14-34 days after a BNT162b2 or mRNA-1273 (Moderna) booster after a ChAdOx1-S (Astrazeneca) and BNT162b2 as a primary course ranged from around 85 to 95%. Absolute VE ranged from 94-97% and was similar in all age groups. Limited waning was seen 10+ weeks after the booster. Against hospitalisation or death absolute effectiveness of a BNT162b2 booster ranged from around 97% to 99% in all age groups irrespective of the primary course with no evidence of waning up to 10 weeks. This study provides real world evidence of significant increased protection from the booster vaccine dose against mild and severe disease irrespective of the primary course.
https://www.nature.com/articles/s41591-022-01699-1
There’s a very long list of studies showing evidence that vaccination reduces rate of infections AND forward transmission, and which suggest mechanisms by which that might happen.
Of course the data aren’t perfect, there are uncertainties, there is waning, and the efficacy varies across variants. But if you have a theoretical premise that vaccines simply can’t prevent infection, and you stick to it no matter what, you can just convince yourself to dismiss any evidence you don’t like.
Overall, 77.4% (3,833 out of 4,950; 95% CI 76.2−78.6) of SARS-CoV-2 infections, 90.9% (748 out of 823; 95% CI 88.7−92.7) of SARS-CoV-2 associated with hospital admission and 98% (102 out of 104; 95% CI 92.5−99.7) of SARS-CoV-2 associated with critical care admission, as well as all baby deaths, occurred in pregnant women who were unvaccinated at the time of COVID-19 diagnosis. Addressing low vaccine uptake rates in pregnant women is imperative to protect the health of women and babies in the ongoing pandemic.
https://www.nature.com/articles/s41591-021-01666-2
Findings Correlation between RNA copy number and IVT was low for all groups. No correlation between IVTs and age or sex was seen. We observed higher RNA genome copies in pre-VOC SARS-CoV-2 compared to Delta, but significantly higher IVTs in Delta infected individuals. In vaccinated vs. unvaccinated Delta infected individuals, RNA genome copies were comparable but vaccinated individuals have significantly lower IVTs, and cleared virus faster. Vaccinated individuals with Omicron infection had comparable IVTs to Delta breakthrough infections.
Interpretation Quantitative IVTs can give detailed insights into virus shedding kinetics. Vaccination was associated with lower infectious titres and faster clearance for Delta, showing that vaccination would also lower transmission risk. Omicron vaccine breakthrough infections did not show elevated IVTs compared to Delta, suggesting that other mechanisms than increase VL contribute to the high infectiousness of Omicron
The link for that last one wouldn’t pass the filter. Googling will turn it up.
Anyway, anyone who’s actually interested can easily find evidence that vaccines reduce infection and transmission
Callaway 2021 does not mention Israel or viral load. Apparently no one out of four authors, an editor, probably 1-3 reviewers, and you bothered to check it. I’m kind of harping on this because I found this scenario to be very common when doing medical research. Often it seemed like I was the first person to carefully read a paper that had been cited hundreds of times.
You cannot measure “viral load” with PCR. I have been saying since literally the beginning that these tests have never been validated. It took until like two weeks ago for something to be published on this, and sure enough they found little correlation with the amount of virus even when limiting themselves to CT under 27:
https://www.medrxiv.org/content/10.1101/2022.01.10.22269010v2
So you can ignore the dozens of papers attempting to measure viral load with PCR, we can discuss that one if you want.
Speaking of Callaway:
Last month, the Ministry of Health in Israel, a country that has one of the world’s highest vaccination rates, released raw data on vaccinations and infections from December 2020 to July 2021. The ministry estimated that vaccine protection against both infection and disease had dropped from above 90% in the early months of its programme to around 40% by late June — a decline that could be due to the effects of the Delta variant.
https://www.nature.com/articles/d41586-021-02158-6
@Anoneuoid You write, “You cannot measure “viral load” with PCR. I have been saying since literally the beginning that these tests have never been validated. It took until like two weeks ago for something to be published on this, and sure enough they found little correlation with the amount of virus even when limiting themselves to CT under 27:
https://www.medrxiv.org/content/10.1101/2022.01.10.22269010v2
“So you can ignore the dozens of papers attempting to measure viral load with PCR, we can discuss that one if you want.”
I take it that “discuss that one” refers to the paper you linked here. In that paper, they find correlations for Delta of somewhat less or somewhat more than 0.6 (Fig. 1). They call that “low”. Is that what you refer to as “little”? I didn’t think that 0.6 was considered a low correlation. (Note that they report its square, which, of course, is considerably lower.)
@ Russ
You can clearly see from the charts that the same CT value is consistent with a 3-4 order of magnitude wide amount of infectious virus. That is what is meant by “low/little correlation” in this context.
Yes, these things operate on a logarithmic scale. So what? It’s still saying that low PCR results (in order of magnitude) suggests low infectious viral load (in orders of magnitude.)
I mean yes, it’s a noisy measurement. But then, what are you saying, the random noise just happened to line up in these nasal swabs to give vaccinated subjects less measured viral load? The presence of noise in the PCR measurement if anything *reduces* your estimate of vaccine efficacy. What about the stuff people are doing with wastewater PCR, are you saying our viral counts just *happen* to correlate with case intensity? What about every other field where PCR is used?
@Zhou Fang Could you please explain to us what it means that “these things operate on a logarithmic scale”? I do always see them presented that way, but I do not know the reasons. Thanks.
I think some of you would benefit from hearing the kind of standard discussion of this paper in TWiV 854. It starts at around 1:04:00:
https://m.youtube.com/watch?v=b1u4yCAoAZI&feature=youtu.be
They get up to some cringe practices like counting asterisks, but that should at least give you a starting point.
This is the type of paper we needed since March 2020, and even now we only have a single one.
@Anoneuoid I looked at some of the video you linked. It did not seem to explain the use of log scales. Mostly, they seemed to just read the paper. I did find one discussion of what was missing in the paper, however, at https://youtu.be/b1u4yCAoAZI?t=4701 , namely, standardization of nasal-swab sampling. This, of course, could account for increased variability (noise) in the results.
Interesting that you referenced that
…while infectious vial load was significantly decreasedin vaccinated patients.”
Because it’s great example of the evidence you insist on ignoring (and the kind they’ve been discussing at TWIV over an extended period of time, suggesting that vaccines do, in fact, reduce infection AND transmission).
Anoneuoid –
And of course…
Vaccines have been shown to tremendously reduce symptomatic SARS-CoV-2 infections.
And
… this study provides strong evidence for higher infectiousness of the Delta VOC as well as a significantly lower infectiousness and a faster clearance of infectious virus in vaccinated individuals.
Your consistency in treating evidence selectively remains impressive – especially since you have no agenda.
It’s an interesting juxtaposition, and difficult for me to understand. Since you have no agenda, why is your treatment of the evidence so selective?
I’m not sure what Zhou Fang meant, but it is plotted using a log scale because the data varies by multiple orders of magnitude. If it was me, I would put non-logged plots in the appendix but that isn’t standard.
Whether you should calculate correlation coefficients on logged or raw data, I don’t know. It seems arbitrary to me, but most of these statistics give an inflated sense of precision anyway.
For example, in this paper the non-vaccinated samples were collected June 26 – August 29 but vaccinated was July 8 – Dec 4. Did someone new start working there who didn’t stick the swab as deep, does the weather or hvac use affect viral load, etc?
That type of systematic error is way more important than statistics that assume random samples.
@Anoneuoid I agree that systematic error, as well as other sources of variability, are very important. That is one reason why I don’t understand your (and the authors’) statement that the correlation is low.
Say you measure ct = 23, and this is consistent with anywhere from 0 – 10^5 focus forming units.
That is not a useful proxy. The correlation coefficient doesn’t even matter as much as the slope.
Also, the correlation they did see may be exaggerated due to lack of normalizing to total mucus/tissue collected. Ie, some swabs just had less total stuff on them than others.
@Zhou Fang Could you please explain to us what it means that “these things operate on a logarithmic scale”? I do always see them presented that way, but I do not know the reasons. Thanks.
The underlying data distribution works logarithmically. For example, PCR is based (approximately) on counting the number of times a sample has to be doubled up to hit a detection threshold. Thus the core measurement is an estimate of the log-quantity, and if it is measured with some uncertainty, the impact is best modelled working logarithmically.
In a similar way, viral load also works on that kind of scale. For example, viruses have a doubling time, and so if you were to plot viruses for a range of people at a range of progressions of the disease, you’ll find that the distribution is extremely long tailed and thus would have to be logged for usual methods (like the correlation) to be theoretically valid.
Thus, a log-log comparison is the appropriate measure in this instance.
@Anoneuoid
“Say you measure ct = 23, and this is consistent with anywhere from 0 – 10^5 focus forming units.
That is not a useful proxy. ”
You realise, Anoneuoid, that this logic is *exactly* the NHST methodology that you keep blasting all the time.
@Anoneuoid You seem to be saying that there are several reasons why the correlation (defined not using r) is low or should be considered low. However, I thought that originally you pointed to this preprint as proving that there was little value in using PCR measurements to predict infectiousness. Now you seem to be saying that due to measurement errors, this preprint is not reliably measuring the relationship, and therefore is not, in fact, establishing your point. Another way to say this is that this preprint, too, falls in the category of not validating the use of PCR for this purpose. That is different from saying that it invalidates its use. What am I missing?
@Zhou Fang Thanks for these points of how these two quantities are measured. Nevertheless, there is the medical/biological question of how relevant the logs are. Can you enlighten me on that?
> @Zhou Fang Thanks for these points of how these two quantities are measured. Nevertheless, there is the medical/biological question of how relevant the logs are. Can you enlighten me on that?
For the final output it depends on the application. Often you do want to backtransform, e.g. if the audience lacks technical skills and thus will misinterpret log graphs. Then again sometimes logs are more natural. But yes, strictly speaking you should base your decisions on log transformations (or alternatively, using a different model specification) on statistical properties, mechanistic explanations, or preferably both. Whatever that other guy says, this is not an arbitrary decision! This is often absolutely critical.
@Zhou Fang I have learned of one case where logs are more informative: Since the doubling time may be quite short, if you want to predict what will happen or how to treat a patient, the log may be more useful because it is linearly connected to time.
For some reason you seem to insist that the CT value can be a useful proxy for amount of infectious virus, even though theoretically it should not be, and the only evidence published also shows it is not.
I am not sure what you are missing, but it is probably better to email the virologists on TWiV and ask them to explain it. I would also like to watch, so let me know if you do it.
@Anoneuoid I never insisted that CT value is “a useful proxy for amount of infectious virus”. I am merely trying to understand why you say it is not. Now you say that “theoretically it should not be”. Could you please explain that? You also now say that “the only evidence published also shows it is not.” I assume here that you are referring to the preprint you linked and which I have been discussing. You say it “shows” this, and my previous comment to you was precisely on this point: How does it show this? You seemed to agree that it suffers from measurement error and confounding.
Because it is common to have mRNA floating around that is not producing infectious virus. And it is also common for mRNA that is associated with infectious virus to be degraded.
It shows the same level of RNA can correspond to a wide range of viral load, thus making it a poor proxy. And the main confound (amount of mucus/tissue sampled) is expected to make this worse.
Really, don’t believe me. Email TWiV and ask them to explain it.
Anoneuoid. Nowhere in that paper’s graphs (I just skimmed those didn’t read the whole thing) did I see anywhere that they plotted log(FFU/ml) vs ct for a selection of patients. That’s the only data I can think of which would convince me that PCR ct is not a good proxy for infectious virus. Having a spread of +- 1 or 2 on a log10 scale isn’t a big deal, the big deal is… is something like a group of people with ct = 32 showing mostly the same number of FFU as a sample of people with a ct of 18?
Did I miss it?
@Daniel
They present it as genome copies/ml which is somehow converted from the Ct value using a standard:
Ref 28 says:
https://www.ncbi.nlm.nih.gov/labs/pmc/articles/PMC6988269
That Ref 8 says:
https://www.nejm.org/doi/10.1056/NEJMoa030747
Welcome to medical research methods sections.
I once saw Richard Epstein give a talk at the U of C, at a conference on the Bill of Rights. He was introduced, and then, Halfway to the podium, noticed he had left his notes at his seat. He turned back for a second, then shrugged, went to the podium, and gave his talk without notes. And it was brilliant, just pouring out in well-organized paragraphs. Maybe a bit of theater , and he had memorized his talk, but if so, it was an effective bit of theater. He a really interesting legal scholar. So it’s sad to see him make ridiculous statements about Covid, and not learn an caution even after making such utterly wrong predictions.
Sorry about the typos … he WAS a really interesting legal scholar, who did not learn ANY caution.
> Clearly, if the vaccines were undeniably as safe and effective as the government claims, there would be no need for any mandate at all.”
The unfalsifiable argument is the mainstay of COVID contrarianism. We know that ivermectin and hydroxychloroquine work because “the medical establishment” doesn’t support their use. We know that vaccines don’t work because Big Pharma makes them. Etc.
As much as I’ve become used to seeing smart people making incredibly bad arguments when they get “motivated,” the pandemic has resulted in a massive amount of next level bad arguments. Base rate fallacy. Simpson’s paradox. Generalizing across context with no consideration of confounding variables. Extrapolating from unrepresentative, convenience sampling. Embedding assumptions about counterfactual assumptions to reach ridiculous states of overconfidence that just coincidentally align with ideological viewpoints (the widespread predictions of “herd immunity” in May of 2020 was truly a spectacular fail).
It goes on and on. So many arguments being presented that display obvious failures of basic science and logic. COVID, , imo, has unleashed a pandemic of terrible reasoning that has appleaed to massive audiences, Joe Rogan, Brett Weinstein, Jordan Peterson… how many people actually believe that billions have been “hypnotized” into getting vaxed? Is this unprecedented, historically – a function of social media?
Joshua:
In addition to all that, universities (including Stanford) are generally left-wing environments, and so stupid right-wing arguments such as Epstein’s stand out for their politics as well as their inanity. Which is good for Hoover and Epstein to the extent that their goal is to get attention and $, not so good if their goal is to not look like fools.
I feel like looking like a fool doesn’t even exist any more. No matter how bad the argument, you can find tons of support on social media. In fact, it’s almost as if there’s a positive correlation. The worse the argument the more support. This idea that billions (of sheeple) have been “hypnotized” to get vaxed (yes, that’s literally the argument and it’s purveyors have been called credible by a commenters right here at statmodeling) is incredibly popular. And the more pushback you get from “mainstream media” or liberal academia or fact checkers the more status you get. That’s how the unfalsifiability mechanism works: the more pushback you get from the outgroup, the more credible are your arguments. Arguments that vaccines work, and that anti-vax messaging is full of shit, only reinforces that belief that the vaccines are a failure.
If Rogan gets booted from Spotify for promolgating vax misinformation, it’s proof that his “truth” is too powerful to be tolerated. If Rogan doesn’t get booted, it’s proof that he’s too powerful to be booted – because he’s got “the truth” behind him.
In the end, his “truth” is unfalsifiable.
If he had integrity at some earlier point in his career, Epstein has chucked it all aside. Here’s a post I wrote back in April 2020 that sort of speaks for itself: https://angrybearblog.com/2020/04/richard-epstein-peak-dishonesty
I’ve wondered a bit about the collapse of honesty and associated scruples in recent years (not universal of course). Here’s another meditation (on shamelessness) that’s relevant to Epstein and his crowd: https://econospeak.blogspot.com/2018/11/the-death-of-shame.html
Regarding vaccine mandates in particular, I think it’s important to separate the real policy questions from the bad behavior of some of the anti-public health crowd. There are two arguments for mandating a vaccine, reduced transmission and reduced demand on limited health care resources (or public subsidies for health care, which are enormous in the US). Initially the case for Covid vaccines was strong on both counts. With omicron, the first justification is greatly reduced (we don’t know quite by how much), and the second plays a greater role.
There *is* a case to be made for regulating personal behavior simply to avoid the public costs of increased hospitalization etc.; an example is seat belt mandates. This is a case where the personal cost of the mandate is very low and the public benefit large. Other instances are less clear. We have restricted tobacco use (and advertising), but we have stopped short of banning cigarettes altogether. What about sugary drinks and foods? A soda tax is probably a good idea, but how much further would we want to go? I think the case for Covid vaccine mandates is now in or near this territory. It is likely to be weaker than the case for indoor mask mandates, restrictions on certain types of gatherings, ventilation, and the provision of paid sick leave. (That’s why the Biden administration’s case before the Supreme Court in NFIB v. DOL was so disingenuous; they had refused to take far more effective steps proposed for an emergency OSHA standard.)
Again, I don’t want to let Epstein and his ilk off the hook. But they may be right for the wrong reasons.
As I mentioned in a post above, please show me where pfizer/moderna ever claimed reduced transmission?
Where did you get this idea? Because it is wrong and known to be basically zero probability beforehand.
Not sure what you mean by this. Reduced transmission, in a community sense, is a combination of reduced susceptibility to infection and reduced probability of transmitting conditional on infection. Are you saying that the mRNA vaccines had no effect on either prior to omicron?
There’s also evidence that the same phenomena exist (although relatively less so) with omicron. Of course, the data are attached to LOT of uncertainty – but they do exist.
I mean that lasting immunity towards this type of virus has never been observed, and that it is well known since the 1960s that intramuscular vaccines do not induce the mucosal immunity required to significantly reduce infection and transmission. This was then verified in the animal trials for the mRNA vaccines.
If you get humoral antibodies really high some will leak into the mucosa for a month or so, but thats it.
There were silly studies published claiming the vaccines reduced transmission but they are all confounded by different testing rates, seasonality, and immunosuppression during the first week that increases the rate of infection (a kind of culling effect).
I remember trying to bet people last spring that Israel was going back over 5k cases/day by the end of the year due to the widespread misinfo about herd immunity.
I see a couple of things going on with Epstein. One is that being very smart can get you in trouble; because you are good at arguing your way out of binds, you get careless about them. The other has to do with difference between law and science (I dealt with both when I worked in the California water wars). To oversimplify, but not too much, in law the categories are primary, and in science the data are primary. A lawyer’s job is to support some position, and smart lawyers like Epstein are great at using categories to do that.
Apropos the first of the two things you mention, Michael Shermer said that the reason smart people believe weird things is that they are good at constructing clever arguments to support the weird things they came to believe for other reasons. Someone who is good at reasoning in general will likely be good at motivated reasoning as well.
Michael Shermer! It’s hard to take that guy seriously after hearing this story.
The beautiful thing about skeptics who are cranks about one special topic is that you can use the teachings of skeptics to understand why they do it. In Shermer’s case, he was clearly motivated to reason why his right-libertarian politics were the only politics a sensible person could have.
I can’t speak to the radio thing or the allegations that he raped a colleague.
Sean:
I know nothing about the politics or the rape allegation. The radio story is just weird because it’s so obviously stupid. Whenever I hear about things like this, I think of the famous story of Arthur Conan Doyle’s belief in those obviously faked photos of fairies. It makes me wonder if the very ridiculousness of the idea is itself a form of commitment device. As I noted in the radio post, it’s accepted for people to have areas of their life that they don’t subject to rational analysis—no one’s gonna get on the case of some scientist who expresses a belief that Moses parted the seas or Jesus rose from the dead or whatever—but somehow Doyle and Shermer cross a line by applying a hard-line anti-rational take to everyday objects.
Andrew, in my case I had already decided to be leery of him before he wrote the weird thing about a ghostly radio (he started writing books about how science proved his politics were the only rational ones). So without reading his original article about the ghostly radio I can’t say what it tells me about him.
Sean,
Fair enough. I’d never really known anything about the guy except that he was some sort of well-known skeptic. Maybe I’d heard about him because he’d written books that were reviewed in the New York Times, something like that? I’d never read anything by him because it had never come up somehow. It’s not like I need to read a book telling me that mermaids aren’t real or whatever. So the first thing I ever read about Shermer (other than the occasional book review, I guess) was the radio story. It’s a pretty good story, and I guess I admire Shermer for sharing it, given that it makes him look pretty foolish. On the other hand, despite my admiration for his openness in sharing the story, I do think it makes him look foolish, and it makes me not so interested in the rest of his story.
I wouldn’t call them clever. Smart but dysrational or dishonest people are often good at constructing fallacious arguments.
Epstein’s audience is the Federalist Society and like-minded judges and politicians. He’ll cite the Epoch Times so they don’t have to, they can cite an eminent legal scholar instead.
Dan:
So you’re saying that the Hoover Institution serves as a kind of laundering operation, where they apply some combination of money and prestige to raise the status of crappy arguments? That’s an interesting idea, and I guess in that case the Hoover trustees could justify their promotion of stupid arguments such as Epstein’s because they further the larger cause of small government or whatever.
On the other hand, reputational inference goes both ways: every time Epstein etc. publish something stupid, this degrades the reputation of the institution. That’s just Newton’s third law.
On the other other hand, maybe the Hoover trustees see this as a plus, in its own way. Dumb things being promoted by Hoover make Hoover look bad . . . but Hoover’s part of Stanford, so that makes Stanford look bad (kinda like this), and Stanford is part of liberal academia.
So every time Hoover publishes something stupid, they do their part by discrediting academia. I’m not saying this is their primary goal or even that it’s high on their list, but at the very least it reduces the disincentive they might otherwise have about promoting asinine arguments.
Speaking in general terms, I think that loyalty toward institutions can be problematic. One issue I have with the leaders of organized psychology is that they’ve been loath (with rare exceptions) to criticize even some of the worst behavior in their field—and I think a lot of this is coming from personal institutional loyalty. In short: they’re psychologists, and they don’t want psychology to look bad. Or: they’re academics, and they don’t want academia to look bad. So it’s kind of refreshing for Hoover to have the attitude that they’re academics but they do want academia to look bad. It allows them to be more open! On the other hand, it also provides a perverse incentive for them to publish crap, as we see in the case of Epstein.
+1
Because most of the participants of this blog know about ordinary transitivity as in
if you pick your three distinct real numbers, you can order them in a certain way such that a<b and b<c. By the axiom, it follows that a<c.
there is a tendency to overlook the possibility that when it comes to citations,
a is related to b and b is related to c means that c is unconnected with a
unless you look carefully.
There is this belief, among some intellectuals, that they can evaluate observational evidence (no matter where this evidence comes from, COVID, Climate Change or anything else) in a dispassionate (ignoring personal political & philosophical beliefs), rational, and competent way. Very frequently this is not the case, and the results are, at the same time, catastrophic and laughable. Epstein is, as you say, a fool, and we should not be surprised at what he has to say. More surprising is the size of his audience and the reputation he developed among conservatives. Even more surprising is what has been said by people who are properly trained, very experienced and who should know better. Even more surprising is what this people have to say about science, democracy, and other big issues; see for example
https://www.nationalreview.com/2021/12/america-needs-a-rebirth-of-science/ and
https://www.tabletmag.com/sections/science/articles/saving-democracy-from-pandemic
+1
> There is this belief, among some intellectuals, that they can evaluate observational evidence (no matter where this evidence comes from, COVID, Climate Change or anything else) in a dispassionate (ignoring personal political & philosophical beliefs), rational, and competent way. Very frequently this is not the case, and the results are, at the same time, catastrophic and laughable.
I agree. It’s remakable to see how many people who clearly have skills and experience in doing scientific analysis flunk basic principles of science and logic when they step outside their domains of expertise. Scientifically-minded people think that Googling and reading abstracts substitutes for conducting actual research. Almost always, they distinguish their own conclusions from those who disagee by stating that they, unlike others, are seeking “truth.”
Again, I think that technology has contributed to this problem. We are pattern-finding machines and massive amounts of data readily available online makes finding any patterns you want incredibly easy.
Basically everyone, no less those with an academic background, thinks they have what it takes to conduct epidemiological analysis of COVID – and they can easily point to a link to prove it!
But it’s not unique to that cohort. I recently listened to a Rogan podcast where he explained he’s an expert on covid beciswe he has a file on his phone where he’s taken a lot of notes.
And it’s not only a phenomenon when people step outside their domain of expertise. Renown epidemiolgists like Ioannidis have produced analysis that breaks the most fundamental precepts of their science (sorry Andrew) and then as a result become a fountainhead for a cascade of armchair epidemiolgists.
Andrew, yes it’s the first stage. Typically, Epstein will be cited in some amicus brief, the amicus will be cited in an obiter dictum footnote, and the footnote will later be cited as stare decisis. A lot of crap law gets smuggled in through dicta in footnotes.
I saw Dan’s observation in action. I worked in a federal regulatory agency a long time ago.
It was well known that the staff would put ideas and concepts into footnotes of agency orders. Such footnotes were often not carefully reviewed at the higher levels in the agency. A year later, the staff could point to these footnotes as a prior policy that should be expanded or more clearly defined.
Bob76
Don’t mandates ultimately fail for the same reason the War on Drugs fails? If you make a law against non-violent behavior such as smoking pot, or not wearing a seatbelt, will I find ways of attaching the seatbelt behind my back, for example? Did drug laws make me into a drug addict?
Speaking of misinformation, why have I not personally suffered “a winter of illness and death”, as Biden promised me in December?
Why do you assume ergodicity, while my personal experience proves you wrong?
Rsm:
It varies. Vaccine mandates work for measles etc. Seatbelt use went way up after seatbelt mandates. Drug laws have effects too. Kids aren’t allowed to smoke in schools.
Regarding “winter of illness and death”: it depends on what is your comparison point. Estimated current excess deaths from covid is 10-20% (see here). That’s a lot of illness and deaths. I looked up Biden’s “winter of illness and death” quote, and he was referring to the unvaccinated. That 10-20% excess death rate is for everyone; it’s got to be much higher among the unvaccinated. So I don’t see that Biden was wrong. That said, sure, it could be worse. Even an excess death rate of 50% isn’t as bad as an excess death rate of 100% or 200% or whatever.
Finally, I don’t understand what you’re referring to with “ergodicity”: is this the hypothesis that if Epstein talks for long enough, eventually he’ll say something correct by accident?
Andrew, doesn’t non-ergodicity mean that my individual experience, despite being unvaccinated, directly and personally contradicts Biden’s promise to me, that I would become sick and die, this winter, due to being unvaccinated?
In other words, if you and Fauci followed me around measuring the virus around me, would your science determine that (having been a germophobe for decades) I am perfectly capable of avoiding infection without needing a vaccine?
Why do vaccine mandates remind me of Selective Service registration in the early 1980s, which I was forced into despite being anti-war? (Are there always better alternatives than war, and forced vaccinations?)
Is it too cognitively dissonant to observe that vaccine mandates are about social control, not virus spread?
> doesn’t non-ergodicity mean that my individual experience …?
Don’t believe everything you read in the internet.
Rsm:
You write, “doesn’t non-ergodicity mean that my individual experience, despite being unvaccinated, directly and personally contradicts Biden’s promise to me, that I would become sick and die, this winter, due to being unvaccinated?”
I have no idea what you’re talking about with ergodicity, but I think you’re misunderstanding statistics. A 10-20% excess death rate is high, and this rate is much higher among the unvaccinated. When Biden said, “We are looking at a winter of severe illness and death for the unvaccinated,” he was not making a personal promise to you, individually, that you “would become sick and die, this winter.” That’s a misunderstanding of statistics, along the lines of the hookah story.
Finally, in your last paragraph you’re using the word “observe” where you should be using “opine.”
Is my personal experience an anecdote that contradicts Biden’s threat?
Do I not get credit for predicting Biden’s statement had no predictive value for me, personally?
If you all were vegetarian germophobe hermits, like me, would there ever have been a pandemic? Have you considered that you might learn how to live healthier lives from my anecdotes, rather than the corrupted anecdotes you are constantly bombarded with?
Why didn’t Biden acknowledge that I, for one, can use non-pharmaceutical interventions and self-isolation to avoid infection without needing a vaccine?
Is it going too far to opine that Biden’s propaganda is as bad as Reagan’s “this is your brain on drugs”?
Must the opinion that researchers are being used as convenient, well-paid tools by politicians eager to extend state control be suppressed, because power is at stake?
Rsm:
You write, “Is my personal experience an anecdote that contradicts Biden’s threat?”
Biden didn’t make a “threat”; he made a statement relating what his public health advisors had told him.
And, no your personal experience of being unvaccinated and not getting severely ill or dying does not contradict Biden’s statement. He did not say that you will become severely ill or die from covid or even that the probability you will be severely ill or die from covid is 50%. He said, “We are looking at a winter of severe illness and death for the unvaccinated.” A death rate of much less than 50% can still be fairly described as “a winter of severe illness and death.”
This is basic statistics. i agree that lots of people misunderstand these issues, even lots of doctors—see for example the hookah story linked to in my comment above—; still, it’s ultimately a simple issue.
I suggest if you want to continue with this, you take it to twitter or 4chan or some other forum that is more suitable for political position-taking and is more forgiving of statistical misconceptions.
“We are looking at a winter of severe illness and death for the unvaccinated.”
Are you claiming Biden was not trying to bully me into getting vaccinated?
Or are you saying it was rational for me to believe Biden’s statistical statement did not apply to me, because I use non-pharmaceutical interventions stringently?
Why do you assume I’m uneducated in statistics, when I worked in market research for a decade, starting as phone surveyor and ending up as an SPSS and SAS programmer preparing statistical tables and charts for client reports (my first boss was a statistics professor)?
Will my opinion that vaccine mandates are really about social control, not virus spread be the excuse you use to ban me here, as I was banned from twitter (not for vaccine talk though)?
Why do vaxxers blame me for virus spread that I’m not contributing to? What illness and death this winter have I caused by being unvaccinated, and if, as I bet you could verify by following me around, none, why pressure me to get vaccinated?
Can I expect a ban in 1…2…3…?
Rsm:
Again, if your goal is to troll, I suggest 4chan; you should be able to get more action there.
If you’re serious:
1. I think Biden was making a public health announcement and I wouldn’t call it bullying at all.
2. The fact that you’re healthy does not invalidate a general statistical statement about risks.
3. I never said that you were uneducated in statistics. As we’ve discussed many times on this blog, statistics is hard, and people who are educated in statistics make statistics mistakes all the time.
As for the rest: believe whatever you want.
I’d like to make a meta-comment about rsm’s style of asking “questions” rather than making claims or stating facts or opinions. I put “questions” in quotes because they’re clearly intended to be rhetorical statements cloaked as questions, rather than genuine questions. There are a couple of long-time commenters here who tend to do the same thing. I suppose some people must find this style of argument persuasive, otherwise they wouldn’t imitate it, but I find it extremely off-putting. Sure, throw in the occasional rhetorical ‘question’ every now and then, it’s fine, I do it myself, but to try to present an entire argument this way just seems passive-aggressive. I normally find it irritating but in this case it is so extreme it is quite funny. “Is it too cognitively dissonant to observe that vaccine mandates are about social control, not virus spread?”, that just cracks me up.
Two things I note about most people who use this style — or at least, those I’ve seen — are that (1) they don’t actually care about the answers to their ‘questions’, and (2) if you do have an answer that doesn’t agree with their argument, they just ignore it and ‘ask’ another ‘question’. We are seeing that play out here in this thread.
In general I try to evaluate arguments on their merits. If you’re defending yourself in court it ideally shouldn’t matter how good or bad your lawyer is, you should win if you’re innocent and lose if you’re guilty. If someone is trying to make a point about politics or public health or whatever else it shouldn’t matter if they argue well or poorly, we should try to evaluate the validity of their point rather than the quality of their argument.
But.
Oh, man, when I see someone who won’t take any kind of stand, they just hide behind a long long list of fake ‘questions’, I really have trouble separating the argument from the arguer. It’s just such a chickenshit way of trying to make a point.
This is good:
https://twitter.com/nirav_mainecdc/status/1487455146208600066?t=lGFfs4g3tiUyEbAFO2LdNg&s=19
I appreciate that the quality of commenters on this blog is high enough that, without a second thought, I click on a link with no other knowledge besides “this is good”, and am suitably rewarded. That was good.
Not completely unrelated to this post but mostly related to how vaccines are/should be perceived. I recently saw two articles (one looking at data from the UK, the other from the US) showing a higher mortality rate in unvaccinated individuals from non-Covid causes:
1) https://www.researchgate.net/publication/356756711_Latest_statistics_on_England_mortality_data_suggest_systematic_mis-categorisation_of_vaccine_status_and_uncertain_effectiveness_of_Covid-19_vaccination
and
2) https://www.cdc.gov/mmwr/volumes/70/wr/mm7043e2.htm
What’s the cause of this? Should we downgrade our estimates of vaccine efficacy in light of these numbers? Or should we ascribe the lower non-Covid mortality in the vaccinated population to the vaccine as well?
I’d also like to see Andrew do a post on this data. The cdc study was discussed in the comments awhile back, it seemed like what they actually did is primarily compare pre-vaccinated to post-vaccinated. The key line is:
I can try to go find that discussion if you want.
Also, all cause mortality in the RCTs was low compared to life tables and about the same (actually ~15% higher in the vaccinated). Overall I have seen no good evidence that there is a net benefit in mortality.
If you could find that discussion, I’d really appreciate it. Thanks!
This and Mendel’s comments above:
https://statmodeling.stat.columbia.edu/2021/10/07/the-covid-wager-results-are-in/#comment-2032087
Thank you.
The paper at the first link is barely coherent.
And why would a higher non-COVID mortality rate for the unvaxxed indicate anything about vaccine efficacy? Here are two straightforward explanations for the pattern:
-COVID deaths miscategorized as non-COVID by anti-vaxxers (some places in the US the medical examiner will take family members at their word about the cause of death)
-Anti-vaxxers are dumber and generally worse at decision-making, which impacts their health/mortality in other ways than COVID.
If you really can’t figure out why, you shouldn’t be calling anyone dumb. It is because even if the vaccine had zero effectiveness it would still appear ~66% effective when comparing deaths in vaccinated vs not.
As mentioned above I don’t think that is what is going on here though. Instead pre-vaccine death rates from non-covid causes were 3x higher than post-vaccine. If so, I’d guess due to less anxiety and unnecessary contact with the healthcare system. But I had to reverse engineer the CDC’s data to get to it.
I don’t know why they can’t just release a simple comparison of all cause mortality rate by age group in vaccinated vs not.
i know your whole thing is tediously derailing threads, but I’d be glad to hear from someone who is *not* an anti-vaxxer whether the statement “It is because even if the vaccine had zero effectiveness it would still appear ~66% effective when comparing deaths in vaccinated vs not” makes any kind of sense or is in any way relevant to my point
@dl
The study only included people who got a flu shot in the last two years:
I’m not sure what your difficulty is in understanding why higher baseline mortality rate would exaggerate apparent vaccine effectiveness, but your proposed explanations about “anti-vaxxers” don’t make any sense either.
Anoneuoid
Your comments spur me to ask 2 questions.
First, you seem to be on a constant tirade that amounts to insisting that COVID has not resulted in net deaths and the vaccinations are completely ineffective (or have not been demonstrated to have any effect) – am I correct about what you are saying? If so, why are you – who appears to be quite smart – insisting on such an extreme position? Sure, much of the analysis is bad, the media reports are almost totally bad, but do you really believe the vaccinations have 0 effect (in the NHST terms I know you detest, you would not reject the null that vaccinations have no effect)?
Second, on your point about what the CDC releases, I have a broader question. I’ve been searching for individual case data – the closest I can find is this: https://data.cdc.gov/Case-Surveillance/COVID-19-Case-Surveillance-Restricted-Access-Detai/mbd7-r32t. There are 2 publicly available data sets and a third that you need to apply for access to. I also looked at the actual case report form that the CDC uses (it is posted on their website). Amazingly, I can’t see that they ask for vaccination status at all. Is that correct? They collect all sorts of information about symptoms, testing, risk behaviors, but not vaccination status? Can this be true?
I would say I’m irritated about how hard it is to get any individual data – some of the fields they hide for privacy reasons seems unnecessary to me – it would be difficult (and a waste of time) to reverse engineer the case report and figure out who the report applied to? In our hyper-politicized world, I guess I can imagine that some people might attempt it and use this personal data for nefarious purposes – even more so if the vaccination status was also part of the report. But since it does not even seem to be collected on the form, it isn’t part of this data reported (clearly they are collecting vaccination status somewhere else, just not on the COVID case report form).
If someone knows these data sources, please correct any misunderstandings I may have.
Nope, just that you need to account for these other factors so we have accurate information. Not pretend they don’t exist.
Anoneuoid *
> higher baseline mortality rate would exaggerate apparent vaccine effectiveness,
Of course. But there are two problems with how you approach that isuse:
(1) whether it’s reasonable to assume that a higher baseline mortality rate among unvaccinated would apply to all the many studies that have shown lower rates of covid death among vaccinated as compared to unvaccinated and,
(2) even assuming that baseline rate discrepancy IS present for ALL those studies, it would a be wide enough discrepancy to fully explain the higher rate from COVID deaths among the unvaccinated.
This is like when you talk about “seasonality,” where you treat the existence of seasonality and myriad other factors potentially affecting infection rates, as if they are somehow mutually exclusive – in other words, as if seasonality in itself explains all trends in infection rates.
It’s also like how you treat evidence that vaccination lowers infection and transmission rates. Again, there, you go from problems with the data – for example controlling for all potential confounding variables – to saying that categorically vaccination doesn’t reduce infection and transmission.
Anoneuoid –
> Not pretend they don’t exist.
Please do complain about straw men once again.
Arguing from personal incredulity and bad faith characterizations is just a bad approach.
As the ONS says, “Changes in non-COVID-19 mortality by vaccination status are largely driven by the changing composition of the vaccination status groups because of the prioritisation of clinically extremely vulnerable and people with underlying health conditions, and differences in timing of vaccination among people who were eligible.”
https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/bulletins/deathsinvolvingcovid19byvaccinationstatusengland/deathsoccurringbetween1januaryand31october2021#monthly-age-standardised-mortality-rates-by-vaccination-status-non-covid-19-deaths
They have a nice graph later on where they show that non-covid vaccinated mortality rises in each age group when vaccination is opened up for people with health conditions and declines when it is opened up for the general population. The paper you linked *got their data from that source*, so the fact the electrical engineers doing that statistical analysis chose to omit all mention of the ONS’s explanation and analysis seems, in the authors’ own words, “extremely disturbing”.
Speaking generally, the fundamental problem is that the individual choice to elect to get vaccinated at certain times, and then change your behaviour in whatever way (before and after) creates a massive confounding problem, and you aren’t going to get to the bottom of that without a waaaaaaay more sophisticated analysis than just pointing and going “oooh, line goes up”.
Worst crossword puzzle I’ve ever seen in my life was in the Epoch Times.
Andrew, I have a challenge for you. You have a very strong anti-conservative bias in all your political writing. Do you actually know how conservatives think? What actually goes on in their minds? How conservatives would describe their own beliefs? Nothing you’ve ever written politically on this blog even remotely hints that you do. It used to be considered normal, even required, to be able to articulate an opponent’s position in words that THEY would agree described their position. Only then could you take your philosophical axe to their arguments. The idea was that only then, after you described the opposing viewpoint accurately, could you be supposed to understand their position. You didn’t have to agree with it, just articulate it well and understand it. My challenge for you is this: take an intellectual conservative (Edmund Burke would be the preferred candidate, but a more modern one would do as well, provided that he or she is a deep thinker and a conservative), find out what they say (certainly NOT what the mainstream media, including the ridiculously left-biased Wikipedia, says they say: primary or sympathetic secondary sources only), and post your summary of what they say (has to be in your own words, not just quotes) on this blog in such a way that I can agree with you: this is substantively what they say. The reason Edmund Burke is preferred is because he was the most learned, articulate, respected conservative of his day, and he essentially began the conservative movement. The basic principles of conservatism that he laid down still serve to motivate conservatives to this day.
Adrian:
Is there something you disagree with in my post above? I assume you don’t think that Epstein was correct in his prediction that the number of coronavirus deaths in the U.S. would max out at 500, or in his statement that, “Clearly, if the vaccines were undeniably as safe and effective as the government claims, there would be no need for any mandate at all.”
I guess what I’m saying is that the above post should not be taken to represent a criticism of conservatism or conservatives. It’s a criticism for Epstein for being unreflective and a criticism of Hoover for promoting his writing.
Well, fair enough if you don’t think of this post as critical of conservatism or conservatives. I think many conservatives would disagree with a number of things in this post, and would likely regard it as an attack on conservatism, even if that’s not what you’re thinking. Many posts of yours are attacks on conservatism (and even conservatives), and include rather denigrating language (sneering and ridicule, such as “What an idiot. Jeez.”) that, to me, seems to promote more heat than light. Why attack people? Nothing wrong with attacking ideas, but I draw a sharp distinction between attacking ideas and attacking the people who hold those ideas. I stand by my challenge: I think you should show that you understand conservatism before you attack it. I think even ideas deserve that much respect. You cannot reason about that which you do not apprehend.
I don’t dispute Epstein being wrong in his predictions, although I do think it’s appropriate at this point to question COVID data – all of it. If PCR tests are unreliable after 35 iterations as we’ve been told, how many COVID tests were iterated over 35 times? Which tests were PCR, and which weren’t? Of the COVID deaths reported, how many were actually caused by COVID versus a comorbidity? Etc.
I assume you’ve never been wrong in a prediction? Are you using his incorrect prediction as a means of discrediting everything he says? If so, would you think it fair if I discredit you in everything you say just because you were wrong about a prediction? I’m wrong about stuff all the time, but that doesn’t mean I’m never right. Surely that goes for most people. Otherwise, you might find yourself on the wrong side of the poisoning-the-well fallacy as well as hasty generalization, and ad hominem for good measure.
I think there is an underlying assumption Epstein makes in his statement that, “Clearly, if the vaccines were undeniably as safe and effective as the government claims, there would be no need for any mandate at all.” That assumption is the Law of Human Action: at any moment in time, given a decision to make, a person will weigh the options available, mentally evolve the outcomes in time to imagine how they turn out, and then choose the option with the best outcome for that person. Surely this Law is beyond dispute. However, I find that people on the left appear to act as if the ordinary Joe Blow is an absolute idiot, never doing what’s obviously (at least, to the leftist) in his own best interest, and therefore he must be forced to do so. This is simply not the case: both the assumption of idiocy and the assumption that he must be forced. Joe Blow is a lot smarter than leftists think. Joe Blow knows a lot more about these vaccines than leftists think. Go talk to an average plumber or electrician or other blue-collar worker. And why should anyone assume knowledge of what’s best for everyone? I deny that. I, Adrian Keister, certainly don’t know the specifics of what’s best for Andrew Gelman. So the central question I have is this: if an action is good to do, under what conditions should it be made mandatory? Under the general principle of individual freedom, I would say hardly ever.
I would fault Epstein in his statement with not including a very important piece of the puzzle: ethical development of vaccines. No one denies that all COVID vaccines on the market were either developed with, tested on, or manufactured with (not mutually exclusive categories) aborted fetal cell lines. The manufacturers themselves say that. And many pro-life people such as myself are very NOT OK with that. I am unvaccinated, and I have no intention of vaccinating myself with any existing COVID vaccine, because of what I regard as the unethical development of said vaccines. I am basing my decision on the facts, not on so-called “misinformation”. So Epstein should have said, “Clearly, if the vaccines were undeniably as safe, effective, and ethically developed as the government claims, there would be no need for any mandate at all.” If you accept the Law of Human Action, and if the evidence is that the COVID vaccines are safe, effective, and ethically developed, who wouldn’t want the vaccine? I would take it! The label “anti-vax” is just so superficially slapped on to anyone who chooses not to take the vaccine, without ever investigating why. The data is ambiguous: I would claim that the COVID vaccines are not safe (see the VAERS system reports, now substantially tampered with by deleting a lot of entries), not very effective (even the CDC doesn’t claim the vaccines prevent infection), and certainly not ethically developed. The label “anti-vax” is also very superfically slapped on to people who are in favor of vaccines in general, even the COVID vaccines in particular, but strongly oppose vaccine mandates. Vaccines and vaccine mandates are VERY different things, just as sharing and being forced to share are VERY different things.
I am only going to dissect one part of this because I have other things to do. Do you not understand excess mortality, or are you unaware of it? We don’t get hundreds of thousands of excess deaths without something significant happening that didn’t happen in other years. The thing that didn’t happen in other years is COVID. The question isn’t “with/of”, it’s “if not for”, and it’s very clear that a million people give or take are dead who would be alive if not for COVID.
I noticed lots of unprecedented things happening, most highly correlated with positive covid tests in an area.
Some may have largely abated, like sending sick people into nursing homes, overdoses of hydroxychloroquine, misuse of ventilators.
Others such as shortages of hospital staff, anxiety, and life/work interruptions are still ongoing.
Adrian:
1. My problem is not with Epstein making a mistake—I make mistakes all the time!—it’s that he doesn’t seem to have even attempted to have learned from his mistake. For more on this issue in this particular case, I refer you to the Rex Douglass discussion linked to in the above post.
2. You write, “I think you should show that you understand conservatism before you attack it.” Sure, but I’m not attacking conservatism; I’m criticizing some particular things that Epstein and Hoover have done. I don’t think there’s anything particularly conservative about Epstein’s practice of making fatuous pronouncements and not learning from his mistakes, and I don’t think there’s anything particularly conservative about Hoover promoting this stuff. Sure, I understand that Epstein has conservative views on other things such as regulation and taxes, but that’s not what I’m taking about here.
3. I agree with you that it’s appropriate to question covid data, and we’ve had lots of discussions of this on the blog. I can see the argument that questioning data is a conservative position; if so, I’m taking the conservative position here.
4. I respect that you think that there should be a very high barrier before an action is made mandatory by the government. I guess it all comes down to the details. Familiar examples are that we don’t have the freedom to drive 100 mph on city streets, endangering the lives of passersby. I think we are in agreement that there is legitimate debate on public health measures. In some cases, the argument for vaccines is clear, and there is broad agreement that it’s a good policy to require kids get vaccinated for a bunch of childhood diseases. For the annual flu shot, on the other hand, I think the consensus is that it would be good if just about everybody got vaccinated, but it doesn’t seem reasonable to require it. Where does coronavirus fit in this spectrum? These are cost-benefit questions, and there is disagreement. If Epstein had said all that, I’d have had no problem. But he didn’t; instead he made an absolutist statement that was just stupid. I think what happened is that he’s practiced at stringing words together in a way that sounds sensible if you don’t look carefully at what he’s saying. And he has no editor, no one at the Hoover Institution saying, “Hey, Richard. Hold your horses!” That’s a bad combination. I have no editor either, but (a) I think I’m more careful than Epstein is, (b) when I mess up, I explore what went wrong, and (c) this is my blog, it’s not an official Hoover Institution product. (Yes, it’s at a Columbia University webpage, but it’s clearly just my take on things; I don’t represent Columbia’s position any more than Dr. Oz does.)
5. The aborted cell lines thing is another story. I’m not sure what my reaction would’ve been if Epstein had made that particular argument in his article. I still would not agree with your reframing (“Clearly, if the vaccines were undeniably as safe, effective, and ethically developed as the government claims, there would be no need for any mandate at all”) because, again, there is general agreement that childhood vaccines are safe, effective, and ethically developed, and these are mandated for the usual public health reasons. Again, you or others could make an argument against childhood vaccines, but that would be taking a pretty extreme position, which implies that the “Clearly” in that sentence is inappropriate.
Incidentally: why don’t I have a reply button next to every post? I can’t reply to the “anon e mouse” comment below, because there is no reply button there.
You’re too dangerous. They’re trying to censor you
Adrian:
There’s a limit to the nesting of comments so as to maintain readability of the threads on the comments section.
Zhou wrote above:
It has nothing to do with NHST at all, so I don’t know where to start. I mean there is no null hypothesis, significance level, or test.
The relationship between the CT value and infectious virus is simply too weak for it to have any practical value.
Your null hypothesis is that quantitative PCR is useless. Your “consistent with anywhere between X and Y” implies a significance level and test. You look at the measure, decide that it’s too big for your decision boundary, and make a decision that it’s useless and you don’t have to think any more.
That’s NHST. Well, it’s a bit worse than NHST because you’re not even showing how you are making the decision in a replicable way, but I’ll let you off on that.
Whereas if you were smart, you’d ask questions like – let’s take this uncertainty and propagate it. It might be big for an individual patient, but suppose I measure, say, 8000 patients with an average CT of 23 and another 8000 with an average CT of 24, what can I reasonably say about their averages and what is my uncertainty in that? The answer might surprise you.
No it isn’t, but if it was I would be testing whether there was zero correlation. I am not doing that at all. There is surely a non-zero correlation, just one of no negligible importance.
Then you’re doing NHST with an interval null.
Maybe this will help: It does not matter how many datapoints were used, that relationship will still not be useful. In NHST, the significance threshold will get crossed eventually.
Not with an interval null it wouldn’t. But the same problems of NHSTs apply.
Like, run through the logic here. On what basis can you say that this relationship is not useful, even when applied to large samples?
A simple example:
summary(exp(3*rnorm(1000)) -> x)
#See how those numbers span a range of >5 orders of magnitude?
summary(exp(3*rnorm(1000) + 1) -> y)
# also >5 orders of magnitude
Well, our measure is clearly useless and there’s no way we can determine that one sample is bigger than the other, right?
mean(log(x))
# [1] 0.1295984
mean(log(y))
# [1] 0.9925007
Not so useless now, is it?
What use are you trying to convey? I already know one value is going to be higher than the other withour any calculation at all.
There is some key piece of information you are missing and its confusing you, because your posts make no sense. Like a student who missed a crucial day of class. I’m really interested to figure it out.
Are you familar with the differences between hypothesis testing, significance testing, and NHST (aka the hybrid)?
> What use are you trying to convey? I already know one value is going to be higher than the other withour any calculation at all.
The use I am trying to convey is whether you can use PCR test data to say if vaccinated people have much lower viral loads or not. This is the specific thing in question.
> There is some key piece of information you are missing and its confusing you, because your posts make no sense. Like a student who missed a crucial day of class. I’m really interested to figure it out.
No, it’s you who is hopelessly confused. What *actually* is wrong with NHST? Because you seem to think it’s either “doing maths is bad” or “you have to make assumptions” or the long run rejection behaviour peculiar to certain types of hypotheses. This is not correct.
> I already know one value is going to be higher than the other withour any calculation at all.
Also this result is only true with sufficiently large sample sizes. So, how did you know that? Suppose I did it with sample size 100, would you “know”?
I’ve never said any of those things. Over and over I say its the strawman null hypothesis. When you test that instead of your research hypothesis it reverses the logic of science. NHST is literally bizarro science.
Meehl 1967 explained it perfectly: https://meehl.umn.edu/sites/meehl.umn.edu/files/files/074theorytestingparadox.pdf
I don’t even criticize significance testing *when used to test the research hypothesis*, which is what Fisher often used it for.
I appreciate the comments. Your quote from the ONS would seem to explain a higher non-covid mortality in the *vaccinated* not the unvaccinated. The note on the ONS website seems to mostly apply to people who hadn’t completed the 2-dose regime. If you look at the “21 days or more after second dose”, mortality for non-Covid causes remains higher for the unvaccinated. When looking at the MMWR paper, the non-covid mortality rate for after 1st dose and after 2nd dose remains much lower than the unvaccinated. Same in study of vaccinated vs. unvaccinated nursing home residents that the MMWR paper cites.
I’m under the impression that the sickest were first in line to get the vaccine. If that’s the case, then the unvaccinated group would tend to be healthier than the vaccinated group. It could also be that the more health conscious are more likely to get the vaccine. So, a priori, it’s not clear to me which composition effect would dominate.
I understand that it’s a complicated question, but saying “composition effects” doesn’t seem to actually explain why the unvaccinated group has a higher non-covid mortality.
This was in response to Zhou Fang.
The composition problem is general. We can only vaguely speculate about how these explanations work. In the case of unvaccinateds having higher mortality, this could be due to people with specific medical conditions preventing them from being vaccinated. Or it could be people who don’t like going to doctors. Or it could be people who are unvaccinated also travelling more and getting hit by buses. It could also vary by geography and social deprivation. Note in addition that in the older UK age groups, the proportion who are not vaccinated is actually really small.
I’m not aiming to give you a complete explanation here. I’m saying that a proper analysis needs to stamp all these factors down, and without doing that it’s very likely to make errors in interpretation. I don’t think papers like the stuff you linked are worth looking at.
FWIW, I think the correct way to handle this stuff is by trial emulation. But it’s a pretty substantial chunk of research work that I don’t expect to see done for many months if not years. Right now I’d focus on RCT data instead of all these population epidemiology stuff. Better small clean data than big dirty data.
Thanks.
Another interesting window into the efficacy of vaccines on infection rates and transmission rates (of course, the one study rule applies).
We conclude that Omicron BA.2 is inherently substantially more transmissible than BA.1, and that it also possesses immune-evasive properties that further reduce the protective effect of vaccination against infection, but do not increase its transmissibility from vaccinated individuals with breakthrough infections.
(Google for the link – it didn’t get past the filter.
Joshua,
First of all, virologists are allergic to this ‘more transmissible’ definition. There is some immune evasion, but it doesn’t make virus more trans. But if we forget that for a moment, reality is that both vaccinated and non-vacc. can pick up and transmit the virus. The difference is that vaccinated are less likely to carry enough of the infectious load (not what you can measure via PCR) for a long time to infect more individuals. So, yes there is some advantage but it is indirect and rather irrelevant. One could claim that vaccinated, on average, transmit more, because many of them feel invincible because of the vaccines. The other confounding issue is that we are running out of immune naive people. Most have either been vaccinated or have natural immunity.
BTW, Omicron or any of these variants are not more or less transmissible or benign. It’s probably human behavior that is completely different worldwide b/w all the variants. Transmission is the least understood part of this pandemic. Virologists have a few mice models using sterile animals and that’s about it. The rest is observational Epidemiology.
There is no way to know the real truth about transmission in the absence of all the measures (masks, distancing, lockdowns, etc.) because the world was immediately altered in order to stop the spread.
Navigator –
As I think we discussed before, my take is that it probably isn’t good idea to reverse engineer the properties of the virus from population-level (observational) epidemiology. And sure, there are a shit ton of confounding factors, and behaviors are inherently very difficult to control for scientifically even in RCTs, let alone in observational epidemiology.
That said, I’m a bit surprised by the level of certainty you expressed in your comment and I think that data patterns
(like transmission rates in vaxed vs. non-vaxed or omicron vs. delta households) in research that can’t scientifically control for behavioral factors shouldn’t just be dismissed. Especially when the same patterns play out across different investigations and different contexts. The information from studies such as that one shouldn’t be considered dispositive but FWIW I do use them to help me get a sense of the probabilities.
Of course, I’m a total non-expert and I’d probably be best off just saying I have no real insight into the various probabilities. But that’s just hard to do. I have an inherent interest to look around and see what’s out there and try to make sense of it as best I can – of course with deference to people who are expert (virologists AND epdiemiolgists in this case).
The UK publishes all-cause mortality comparisons by vaccination status. The data is age-standardized. Navigate to “Deaths occurring between 1 January and 31 October 2021 edition of this dataset” and click on the xls file. Refer to the worksheet labeled Table 3 and form your own conclusions.
https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/datasets/deathsbyvaccinationstatusengland
The other worksheets are also fairly interesting to navigate. No, they don’t control for baseline health, ethnicity, and other factors.
For the US, Washington produces estimates of age-stratified case rates by vaccination status, hospitalization by vaccination status, and COVID-specific death by vaccination status.
https://www.doh.wa.gov/Portals/1/Documents/1600/coronavirus/data-tables/421-010-CasesInNotFullyVaccinated.pdf
To respond again to Dale Lehman here:
https://statmodeling.stat.columbia.edu/2022/01/29/fools-at-the-hoover-institution-the-gift-that-keeps-on-giving/#comment-2043990
Here is an example of what I consider to be good science, regarding the pioneer anomaly:
https://arxiv.org/abs/1204.2507
Look at how they compare a quantitative model to the observations and go way out of their way to account for every conceivable source of systematic error. It is beautiful.
Compare to someone like anon e mouse (I don’t mean to pick on them specifically, this is very common) who says above:
There is no effort at all to account for error, nor any kind of quantitative model.
And yes, the root of this type of thinking is NHST. Because in NHST a strawman is tested rather than your hypothesis. So you are rewarded with “significance” for failing to account for sources of error. The incentives are the reverse of what they should be, as a result people are actually *trained* to think in this bizarro science way.
This is described well by the Meehl 1967 paper I also linked in a comment above.
Look, anon e mouse surely hasn’t put in the work of the Pioneer anomaly people to make a mathematical model etc, but they have a perfectly fine informal model that goes like this:
“If a novel coronavirus that infects the respiratory system and other bodily systems and causes damage to the body spreads around the world, then a much larger than normal number of people will die in the following months than in an otherwise normal year”
They then look at the real data, and sure enough, the prediction plays out, including confirmed deaths from COVID and a large number of excess deaths as well.
Nothing more is needed to dissuade them from the assumption “everything was normal and nothing strange happened” because the magnitude of the deviation from that was large, and was exactly what was predicted by the basic qualitative model for COVID… So the posterior probability that “COVID caused a lot of deaths” increased, and the posterior probability for “in fact nothing much happened” went to near zero. This isn’t NHST, it’s qualitative bayesian-style model comparison.
If you want to unseat the “covid caused a lot of deaths” model, you have to come up with a different model which predicts the actual data much better, or alternatively, has a more plausible prior.
+1 to Daniel’s incredibly well-articulated response.
Daniel, your post does not address accounting different factors in the deaths at all. So I don’t see how it is a response to my comment.
Eg, it is well known that early on they were doing “early intubation” and treating the patients incorrectly based on anonymous rumors out of China. This stopped once Cameron Kyle Sidell turned to social media and found agreement amongst many other doctors (hospital administration wanted to keep following the bad guidelines): https://vimeo.com/user49723187
We have never seen anything close to those mortality rates since that practice largely stopped. But we also have never seen an accounting of it either. If this type of thing is not addressed it is going to happen again in the future.
That we might have had lower mortality problems if we’d known more and did things better is not a reason to say “COVID didn’t cause these deaths”… Yes COVID + X caused those deaths and maybe COVID without X would have been better, but that’s not what we got.
It’s like saying “Joe didn’t drown, I mean after all, if he’d been wearing a life jacket he’d have been fine”
In that case we are talking ~20k NYC deaths attributed to covid in Mar-April 2020, and ~20k deaths since.
So it could be a difference as big as ~10k vs ~1k per month due to the wrong response. Ie, an order of magnitude difference. Imagine if they were still following the same wrong protocols.
Calling all of those “due to covid” without some sort of asterisk is misleading. As I said, ignoring these other factors makes it more likely to happen again in the future. The point of collecting this data is to make good decisions. Just saying there were x covid deaths is not helpful.
Anoneuoid
Your logic is flawed. Even if I accepted the notion that many or most of the deaths were due to improper treatment (which I don’t accept), that doesn’t mean they are not COVID caused. COVID is/was a novel disease and treatment will be imperfect – the fact that, in hindsight, we know how to treat it better does not mean that COVID is not responsible. If that logic worked, then a good murder defense would be: “Yes, I shot him, but if he hadn’t pulled out a gun, I wouldn’t have done it, so I’m not responsible.” And it is unfortunate that we have come close to such defenses of late.
I also like Daniel’s response to you. I didn’t quite understand the point of your Pioneer study. Yes, you found a good study. Perhaps it is even better than any other study. But you appear to be dismissing all other studies because you don’t find them as good as the one you like. Strange logic.
I am really hoping that you or someone else will respond to my other question about the CDC case report forms. Is it true that they don’t collect data on vaccination status? I’ve sent that question to the CDC but haven’t gotten a response yet. But the form they use certainly doesn’t have any question about vaccination status.
Say someone tests positive then as a result they get shot with an anti-aircraft gun as supposedly happened early on in North Korea. Would you count the death as “due to covid”? How is that different than a positive test and reduced spO2 so you get prematurely put on a ventilator?
The point is your definition is too vague to be used to make useful decisions. Literally everything that happened prior in the persons life contributed to the death in some way.
How about break the deaths down by death with covid and A, covid and B, etc? Would that be acceptable?
Yes, I would prefer the deaths be broken down by many things, including what type of treatment. I especially would like to see them broken down by vaccination status (as I allude to above). Perhaps that is one thing we can agree on – more detailed data is needed. I could do without the North Korean anti-aircraft guns, however.
There may be a more recent update, but from wikipedia I got this:
https://cdn.who.int/media/docs/default-source/searo/whe/coronavirus19/sear-weekly-reports/searo-weekly-situation-report-33-2021.pdf
So it appears to work. They shot the first guy and then never found any more cases/deaths. Or maybe there are other things going on besides spread of covid that affects the count of covid deaths.
We should not be ignoring these other factors and just citing one number that is influenced by all sorts of things. It is not helpful.
The point was accounting for all various factors is important. I explicitly said this.
I have no idea. But I am sure a transparent look at the data pipeline would be very interesting. I’d expect to find things going on I would never have thought of without direct experience.
You want to throw out key prior information, and I’m the naive statistician here? Back of the envelope, if you added up the maximum theoretical effects of every single hypothesis you’ve ever thrown out here about Covid and assumed every single one were true and had that hypothetical strongest effect on mortality, how many excess deaths could you account for? I literally cannot conceive of a way you could even get to half. And of course there’s zero chance that every one of these hypotheses has a major effect on mortality in reality. So, given that, what actually changes about the vaccine or NPI conversations if we account for your uncertainty? As best I can tell, nothing. If you think you can account for a huge fraction of excess mortality through some mechanism other than “Covid directly killed most of these people”, show your work! Put a number on it. You never do, though.
Or, put more calmly, what Daniel said while I was typing my response.
Here is another source of error when attributing covid deaths not mentioned above, hospital acquired infections:
https://www.ncbi.nlm.nih.gov/labs/pmc/articles/PMC8026275/
Did those patients really need to be in the hospital where they risked a co-infection that could double their chance of dying? How many were there because of anxiety/hysteria, and otherwise would have roughed it out at home?
I don’t know (probably no one does), but this is the type of accounting we need to see. Assuming everyone who tested positive then died is due to covid does not work. And often there are multiple factors required. A seed resting on asphalt is not likely to grow into a plant, while one planted in fertile soil is. Both the seed and the soil are required.
To end with an anecdote, about a month ago my friend tested covid+ and noticed shortness of breath when moving around the house. He was concerned so went to the hospital where they took a chest xray and did whatever tests, then they had him go to the waiting room filled with coughing people for 7 hours before sending him home. He never got any results or advice. Eventually he got better, but the visit was only a risk. There was no benefit.
Nobody assumes “everyone who tested positive then died is due to covid.” Here https://www.cdc.gov/nchs/data/nvss/coronavirus/cause-of-death-data-quality.pdf is a CDC pamphlet on this. “When COVID-19 is reported as a cause of death on the death certificate, it is coded and counted as a death due to COVID-19. COVID-19 should not be reported on the death certificate if it did not cause or contribute to the death.”
Your argument (upthread) that if someone dies of COVID who would have survived if they had gotten better care, they shouldn’t be classified as a “covid death”, seems ridiculous to me and to the other commenters who have posted here. Indeed, even if someone received no care at all I’d call them a “COVID death” if they died because of health complications induced by COVID. Yes, yes, there are gray areas etc. etc., but you are going to get no traction with your argument. A word means whatever most people think it means, and most people think “COVID death” includes people who catch COVID, develop severe cough and lung congestion because of it, and eventually die of lung failure, even if those people could have survived if they had gotten optimal care. And no, nobody thinks people who die because they’re shot should be counted, please stop being a dick about this stuff.
As to your broader point about needing a “broader accounting”, sure, nobody is arguing to the contrary. The map is not the territory. Our description of the pandemic is not the pandemic. If people only look at the top-line numbers — deaths, hospitalizations, cases, etc. — they are going to have a terrible map. We all know this.
As for that anecdote, I don’t know what it’s supposed to indicate. That health care can be lousy? That it’s easy to get sick at a hospital? That it takes a long time to get care? Hey, I have an anecdote of my own, I’m sure you’ll find it very insightful: A few years ago I fell on my bike and gashed my elbow badly enough I needed stitches. I went to the ER, which was not especially full, and waited a long time to get care. For the first two hours I shrugged off their offers of ibuprofen, but after almost three hours I said yes I would like some, and it took half an hour just to get a couple of ibuprofen tabs! In the end it took me five hours just to get some stitches in my elbow. Isn’t that a great anecdote? All the insight I’ve provided there. Says everything you need to know, doesn’t it?!
Oh, I have an anecdote for Anonnyd00d too. Went to a wedding two months ago–about 150 people. Reception was indoors. Half the wedding party was from out of town (Alabama, specifically) and we were told ahead of time that many of those people were not vaccinated. The reception was held in a fancy hotel in a big city. The wedding reception turned out to be a super-spreader event. About 20% of the attendees got COVID! At our table of 10, 5 people got COVID!! Everyone at our table was maskless except my wedding date, me, and another fellow at the table. (Everyone at the wedding reception was maskless except my wedding date, me, another fellow at our table, and an 85-year-old woman.) I am also vaccinated, my wedding date was vaccinated, and the other fellow at the table was vaccinated (this came up during discussion at cocktail hour–a conversation we had while wearing masks). I didn’t get COVID, my wedding date didn’t get COVID, and the other fellow wearing a mask didn’t get COVID. Amazing! Obvious confounding between NPIs and vaccination, but hey, there’s an anecdote for you. I recommend wearing a mask and getting vaccinated. It’s always nice when these anecdotes are consistent with general scientific evidence.
My friend was vaccinated. The post is about risk of medical errors and nosocomial infections.
Why is it so difficult to read the comment and respond to it rather than arguing with some strawman in their head?
reply to Anoneuoid
I should probably give up as you are like a dog with a bone. But I’ll give it one more try. You say that your anecdote is about medical errors. There are plenty of such anecdotes, both COVID related and otherwise. And I may even agree that their are systemic COVID medical errors and that these can pollute the COVID data. I would even go so far as to say that the degree to which they pollute the data is not known and that we may not even be collecting the data needed to quantify these errors with any precision.
But it seems to me that there is a great leap required to go from those observations to declaring that we have no evidence that COVID has led to increased all-cause mortality and no evidence that vaccines have any effectiveness. Aren’t you just using the imperfection of the world and the data to say that we know nothing at all? Wouldn’t that apply to virtually every statistical study ever conducted? You object to straw many arguments, but it seems like this is the biggest straw man of all – since we haven’t included all possible variables in our analysis, the analysis has no value?
Dale,
I have repeated over and over that what I want is for these other causes of death (ventilators, hcq overdoses, hospital acquired infections, etc) to be accounted for. The purpose is to learn from mistakes so lives in the future are saved.
I can really care less if you count everyone who dies after testing positive as a “covid death”, but there is nothing to learn from this number when defined like that. There is no action to be taken based on it.
So call them death due to covid + A, or covid + B, etc. That is fine with me. Then we can move forward to learning from the data.
Anon,
You say “I can really care less if you count everyone who dies after testing positive as a “covid death”, but there is nothing to learn from this number when defined like that. There is no action to be taken based on it.”
Why is it so Why is it so difficult to read comments and respond to it rather than arguing with some strawman? It is not true that “everyone who dies after testing positive” is counted as a “covid death.” Why do you keep repeating this lie? It especially doesn’t make sense to do this because everyone know you’re lying.
Anoneuoid –
My reaction is to what seems to me as a sense of entitlement.
Yes, no doubt errors occur. Systemic errors occur. At the systemic an individual levels there’s incompetence – I’m sure we’ve encountered it many times. As simple and banal a modification as making lists can lead to dramatic improvement. People die because of those errors and incompetence. We’ve probably all even encountered indifference to errors. There’s groupthink and resistance to change.
But the world isn’t binary. There’s no such place as Shangri-la. Unintended consequences are just part of life.
And there really are heroes out there. There are people who dedicate their lives to helping others. Sometimes at significant risk to themselves. Often there is a substantial level of cost. All in an environment where often the system doesn’t have sufficient or needed resources. Sometimes, notably recently with covid, depite vitriol and hostility because of (in my view) misdirected ideological antipathy.
You talk of your first hand experience which has led you to this critical stance you take. But i suspect that if you were in the front lines of providing direct care, you’d at least be somewhat more circumspect, with regard to what comes across to me as a sense of superiority.
Perhaps I’m wrong. Maybe you have spent significant time providing direct care. Or maybe you could try volunteering somewhere to help provide care, as an experiment, to see if it shifts your attitude at all.
Hypothetical:
We know that elevated PSA levels are a marker for potential prostate cancer. We also know that biopsies are often done to better ascertain whether cancer is present or not. We also know that biopsies often miss cancers that are there – and often have bad side effects. We also have some idea of the death rates from prostate cancer. But we don’t know, when someone dies whether it was caused by prostate cancer, side effects of biopsies (including stress thereof), or myriad other things, including surgical mistakes, misread radiology reports, etc. Does this mean that no actions can be taken based on the data we have? Because it is imperfect?
Believe me, the data we have is inadequate. But it is not nothing and it is not a justification for taking no action other than asking for more data. I think your position (Anoneuoid) really reveals that you have a strong prior that COVID is not causing excess deaths and that vaccinations do not work. But rather than present that belief, you couch it in terms of the inadequacy of the data we do have.
This is similar to some of the “analyses” done after the presidential election. Because fraud was/is a possibility, and because the data is not perfect, some people argued that the results could not be trusted – at all. While there were plenty of reasons to be unhappy with the quality of the data, it was not worthless. The lack of perfect data was not a justification for the ridiculous positions taken by some – such as showing that if the distribution of mail in votes was the same as that for votes cast in person, then the probability of the actual mail in vote being as lopsided as it was, was 1 in a quadrillion (or whatever number they actually computed). Of course, there was no data that could definitively prove what the distribution of “true” mail in votes looked like, since we only have data on the mail in votes cast – which would include any fraudulent votes. But the people that made these ridiculous arguments only revealed that they had a prior belief that the election was rigged. I get the feeling that you “know” the answers and that is why you take such an extreme position.
I got back to my original comment to you – why do you need to take such an extreme position? I am in agreement about the inadequacies of the data, but there are enough disparate kinds of evidence that there is “something” to act on, rather than “nothing.”
Anoneuoid –
> Then we can move forward to learning from the data.
Since this is an area of focus for you, then surely you know that there have been endeavors to go over the death counts carefully to examine the differences between “died with” and “died from” covid. There have been, of course, many efforts to account for nocosomal infections and the costs of medical error.
Surely, all of those efforts should be subjected to investigation and critique – but it’s just not like only you cares that these issues be accounted for.
And on the other hand, there are also error factors which have likely led towards undercounting deaths “from” covid – particularly early on in the pandemic and in places like India. I may have missed it, but I don’t recall you posting any comments on that topic.
Is that because I’ve missed it, or if not maybe it’s because you’re more focused on the counting problems in the one direction than the other. If the reason is the latter – then why is your focus selective on that way? Would it be because you have evidence that the magnitude of the error is dramatically disproportionate so as to make undercounting effectively, relatively unimportant?
Sorry. I meant it in the way of “even if you want to define it like that I still wouldn’t care”. It is irrelevant to me, and I have no idea what you would do with this number. What I want is:
1) Odds of dying if you stay at home
2) Odds of dying if you get “early intubation” (seems to be about 5-10x higher)
3) Odds of getting a nosocomial infection and then dying (seems to be about 1.05x higher)
4) Odds of dying if you get a megadose of hydroxychloroquine (seems to be about 1.15x higher)
5) Odds of dying if your family is not allowed to visit you in the hospital (NA)
6) Odds of some kind of medical error and dying from that (NA)
7) Odds of dying if you recently lost your job or had some kind of personal problem due to the live disruptions. (NA)
Etc, I am sure we can think of many more.
> nocosomal infections.
Not to mention nosocomial infections….
@Dale:
Nope, why do you think that even though I have said otherwise multiple times?
As I have repeated over and over, it is that there are other significant factors that need to be accounted for. The presence of the virus is only one thing out of many that changed.
Is this 5%, 10%, 50%? Well that is what we need to find out. Why is there such resistance to this idea? It is just good science to account for as much as you can, like demonstrated in the pioneer anomaly paper.
Anoneuoid –
> 7) Odds of dying if you recently lost your job or had some kind of personal problem due to the live disruptions.
I guessing by this you’re referring to measuring the downside “cost” of NPIs (happy to be corrected if wrong). But of course the odds you’re looking to measure are pretty complicated as you’d have to account for the level of “life disruptions” thar would have occurred had NPIs not been implemented. IOW, you’d have to measure what would have happened had unemployment benefits not been extended, had there not been eviction moritoriums, etc.
Your other curiosity about odds seem to me to be similarly complex; for example, along with counterproductive treatments that might have led to unnecessary deaths there were necessarily lessons leaned that led to improved outcomes – lessons which would likely not have not been learned had mistakes not been made. But maybe you’ve got a handle on all of this – so I do look forward to seeing your results once you’ve made some progress in your research.
Here they found 20% of excess deaths were not attributed to covid in 2020:
https://www.ncbi.nlm.nih.gov/labs/pmc/articles/PMC8136644/
I didn’t closely look at the methods, but we can take this as a starting point.
As they say, a substantial fraction of these could have involved covid in some way. That certainly could be the case but we can’t tell from this data.
However, we *could* also interpret that as a baseline for the effects of NPI’s, stress, etc. In that case, covid mortality rate is also 20% higher than it otherwise would be due to the same factors. Thus 40% of excess deaths can be attributed to the response, rather than virus per se. This would then be further increased by various covid-specific factors (bad treatment guidelines, etc) mentioned above.
So it is certainly possible that excess deaths would be 50+% lower if the response had been different. But, that is impossible to tell from the data available. The reality likely lies somewhere in between the two possibilities.
> The reality likely lies somewhere in between the two possibilities.
Only because you categorically “ignore” the possibility that deaths could have been higher absent the NPIs.
It’s quite remarkable how many people I see doing that, without even bothering to make an actual argument
I’ll respond to this one because it is a good point.
Following the line of reasoning used above this amounts to saying the NPIs increased overall 2020 mortality by ~200k, but then also reduced the covid mortality by some unknown amount. Maybe as high as 200k? That is what would be required for no net benefit.
There is no clear evidence either way (and never will be), but what is the max plausible value?
Anoneuoid –
> Following the line of reasoning used above this amounts to saying the NPIs increased overall 2020 mortality by ~200k, but then also reduced the covid mortality by some unknown amount.
That doesn’t really take into account the counterfactual assumption that you, like so many, wrongly think you can just assume.
Because it doesn’t really take into account what would have happened absent the interventions. Looking beyond COVID as the immediately proximal cause of death – for just one example, what would have happened absent extended unemployment – particularly for someone not going into work because they were concerned about being infected while caring for a senior at home, and getting fired not layed off, and not being able to collect ANY unemployments as a result. The countefactual assumptions are enormously far-reaching. What a out all the illness? The hospitalizations? The impact on Healthcare workers? The costs of treatment? The indirect effects in kids of closing schools?
> There is no clear evidence either way (and never will be), but what is the max plausible value?
I have absolutely no idea what the max plausible value might be, or even the direction of the actual effect.
And I don’t think anyone else does either but fine, if people want to make a careful argument they should go for it. It would be interesting to see. But have yet to see anyone do so. All I have seen is that a lot of people make assumptions, as hi have done – always, of course, in line with their ideological predisposition, or at least to confirm a particular bias.
Consider the following scenario:
There’s a virus that’s 100% lethal that infects 100 people. You have a therapeutic, it it causes a fatal reaction in 10% of the people who are treated with it.
Would you say the treatment killed 10 people or saved 90?.
And then mix in all manner of other complications. Non-fatal side-effects from the treatment. And long-term sequelae from the infection – all manner of “costs” that you can’t even evaluate for years. And then there might be factors like the death from the virus is long and drawn out, or the death from the treatment is incredibly painful. Or you have to factor in age stratification (say the virus is only fatal to children, or primarily to people with comorbidities). Or you have to factor in SES or race/ethnicity, or access to healthcare. And then there’s the risk of what would happen if people left untreated could infect others. The complexities go on and on.
So sure, ask questions about the quality of the data. That’s important. But I suggest that you revisit whether or not you’re cherry-picking. At least it sure looks to me like you are in a VERY obvious way, and that you have been doing so for months. Just like you have with the question of “seasonality.”
Look into the concept of a Spielraum:
https://www.barelysignificant.com/post/corroboration2/
That is what I’m doing.
Oops…laid off…
I hate it when I do that. Can’t even blame it on auto-correct.