Who ever said that every post had to do with statistical modeling, causal inference or social science?
(Above photo sent in by Zad.)
Who ever said that every post had to do with statistical modeling, causal inference or social science?
(Above photo sent in by Zad.)
No face masks? Love the hat. Seems to be a sort of heirarchical modeling to me.
…I want to know who did the nasopharyngeal swabs on that Bronx zoo tiger that apparently has Covid now!
Apparently she didn’t cooperate so they had to knock her out :-)
“The tiger was the only animal tested because the procedure involved general anesthesia, the Agriculture Department said. While there were other tigers and lions showing similar symptoms, the veterinarian wanted to limit the potential risks of general anesthesia to one animal, it said.”
https://www.nytimes.com/2020/04/06/nyregion/bronx-zoo-tiger-coronavirus.html
They have carefully positioned themselves 1.5 cat meters apart.
A topical quote for you, both for the times and the usual subject matter of this blog:
Peter Navarro on his disagreement with Anthony Fauci on the evidence for therapeutic use of hydroxychloroquine for COVID-19. “My qualifications, in terms of looking at the science, is that I’m a social scientist. I have a Ph.D. And I understand how to read statistical studies, whether it’s in medicine, the law, economics or whatever.”
https://www.politico.com/news/2020/04/06/navarro-fauci-feud-coronavirus-treatment-167855
Dalton:
I can’t imagine that Navarro can be saying that being a Ph.D. social scientist gives him expertise or qualifications on reading statistical studies; after all, Brain Wansink, Susan Fiske, Satoshi Kanazawa, etc etc. are Ph.D. social scientists too, and we wouldn’t even trust them to evaluate claims within their own subfields, let alone more generally.
Perhaps Navarro is making a reverse statement: that just as he, Navarro, can have a Ph.D. and be wrong, similarly others with Ph.D.’s and M.D.’s can be wrong too. That is, Navarro’s point is that, if qualifications are what you’re after, there are a million people with doctorates in relevant fields, so we can’t think that just because someone has the degree, it means that we can take them seriously.
I do share with Novarro an annoyance with a traditional deference to M.D.’s. For example, the linked news article says, “Although Navarro has no medical experience, he went on to assert that ‘doctors disagree about things all the time,’ and forcefully defended his credentials as sufficient for him to weigh in on the scientific deliberation over the drug.” But I don’t know how relevant it is that he “has no medical experience.” Dr. Oz has lots of medical experience, and where does it get him? Doctors do lots of great things, but it’s not at all clear to me that medical school and medical practice gives people any special ability to evaluate medical studies.
Is it still April 1st over there?
Fauci makes a measured statement and Navarro whines like a baby.
Maybe there is a reason people are deferential to MDs.
NIck:
I guess I’m saying that Navarro’s argument is self-reinforcing. To the extent that Navarro comes off as foolish or corrupt, this just reinforces his implicit point that you shouldn’t automatically take someone seriously, even if he happens to have an advanced degree in a relevant field.
Who do you trust more, an epidemiologist,with a Ph.D,(not Fauci) or Peter Navarro who “has published peer-reviewed economics research on energy policy, charity, deregulation and the economics of trash collection”.
From what I have read of medical schools, they probably do not teach how to evaluate medical studies the way a good research grad course does but Fauci seems to have spent a lot of his professional life in relevant research posts. He probably has learned something from 40 or 50 years on the job working with top-notch researchers. If nothing else he may have learned to listen to his experts and not some political appointee with no qualifications
Jrkrideau:
Yes, I agree.
For what it’s worth, “In a 2019 analysis of Google Scholar citations, Dr. Fauci ranked as the 41st most highly cited researcher of all time. According to the Web of Science, he ranked 8th out of more than 2.2 million authors in the field of immunology by total citation count between 1980 and January 2019.”
The prestige fallacy trap not-withstanding, I’d say that Dr. Fauci not only learned something working with top-notch researchers, he is a top-notch researcher, probably on one of the toppiest notches.
Wait a sec… Being a social scientist doesn’t inoculate him from being wrong, but surely it gives him the tools to be right. He might have no ability to use these tools, but he is being criticized for not being a medical professional. That’s not a relevant criticism. Here’s the right way to proceed. Let him put his analysis out… if it sucks, it isn’t because he isn’t a doctor. If it doesn’t suck, it isn’t because it isn’t an RCT.
Jonathan:
I agree 100% that it’s inappropriate for people to criticize Navarro for not being a medical professional. As you say, being a medical professional isn’t really much of a qualification at all for evaluating statistical evidence and medicine.
But I disagree that being a social scientist gives Navarro “the tools to be right.” Being a social scientist could give him the tools to be right; alternatively, it could be giving him the tools to be overconfidence, in the manner of Kanazawa, Fiske, Tol, etc.
I wasn’t in Navarro’s class at Harvard (or in any class at Harvard for that matter) and it is true that social scientists can be overconfident, particularly Harvard social scientists. But I don’t think anyone is *taught* to be overconfident, at least not directly. I must have missed the overconfidence classes at Yale, but I sure took a lot of statistic analysis classes, much of it with “anecdotal” data.
Jonathan:
I think a lot of social scientists are taught in graduate school to be overconfident. They are taught to produce research in the style of existing published and celebrated articles, they’re taught to find statistically significant differences and label them as discoveries and to find non-statistically-significant differences and label them as null effects. Maybe some of the people we discuss on the blog figure out how to be overconfident all on their own, but I think a lot of them are trained to be that way.
You may be right about that. But there is another grad school tradition of running down every lead and being extraordinarily cautious to guard your reputation. It is the personality that takes these two traditions and generates either caution or overconfidence. I take your point about NHST generating *any* confidence. That’s a training mistake. But this isn’t about NHST… the clinicians use it too!
All that said, the claim that data is uninterpretable unless it comes from a clinical trial (a clinical trial which will, by the way, invariably use NHST to draw conclusions, alas) is just an error. If you believed it, then none of the epidemiological modeling work would be valid either!
Jonathan:
Overconfidence may be a state of mind or a personality trait. But in grad school, these people learn the tools that allow them to be overconfident in an effective way that can get them publications etc. Kanazawa, Tol, etc. without graduate training would just be ranters. The methods training allows them to do the cargo-cult science thing effectively.
To get back to the earlier discussion: getting a Ph.D. in social science is fine—I’ve been training social science Ph.D. students for a long time—but if you want to count that training as providing “the tools to be right” about quantitative research claims, then you should also count it as providing the tools to be overconfident.
“I think a lot of social scientists are taught in graduate school to be overconfident. ”
Even undergraduate school. A recent comment on another thread discussed the delicacy of the social science undergraduate ego and outlined a solution to this purported problem that, in my opinion, is virtually certain to produce students with low competence and high confidence.
I think teaching overconfidence is implicit, it happens in methods courses where they say essentially “here are the steps to follow to analyze whatever data you might have collected, and if you can get p less than 0.05 and tell a plausible post-hoc story you can be relatively sure you’re correct because the probability you’re wrong is 0.05”
and then people go off confident that they know how to do research…. over-confident because they’re 100% wrong.
the Cartesian dream… Zzz
Great discussion.
Education and the degrees that come with it *can* give a person the tools to be highly competent, but it’s no guarantee.
However, when it comes to a given issue, the issue needs to be decided on the merits of the proposed actions, not the degree or background or experience of the people who proposed those actions.
> However, when it comes to a given issue, the issue needs to be decided on the merits of the proposed actions, not the degree or background or experience of the people who proposed those actions
And this is where the fact that there are more high school dropouts in the US than there are people with a Masters degree runs headlong into every public discussion… (note: I haven’t verified that claim, it was a thing I read recently in the news)
The fact seems to be, a huge fraction of the US doesn’t know enough to make heads or tails of the evidence on something like hydroxychloroquine, even if they had access to that evidence. So the US is one big shitshow of “my expert says this” and “but my expert says that neener neener”.
“The fact seems to be, a huge fraction of the US doesn’t know enough to make heads or tails of the evidence on something like hydroxychloroquine ”
Maybe. But a lot of people knew enough to be wearing masks when Cass Sunstein said there was nothing to worry about. OTOH, people in Iran killed themselves ingesting poison they thought would protect them from the virus. Probably people with Master’s degrees wouldn’t have done that. So I guess when you don’t know the answer, it’s hard to say what it might be.
I agree completely. When someone asks Navarro “why should we listen to you instead of Fauci” his answer ought to be “here is the data and here is what I have done with it. Criticize that all you like. If you aren’t comfortable criticizing the analysis, my degrees are entirely irrelevant.” His answer to the question was clearly not aimed at someone ready to look at the data, but at someone making a credentialing point… He might well be a terrible analysis of data, but the data and his analysis will demonstrate that, not his resume.
Using credential to defend your argument or attack others is stupid. Similarly, one of my pet peeves is people using conflict of interest real or perceived to dismiss a study.
No sensible person would choose to use observational data if RCT data was available. Epidemiological modelling is only used because an RCT is impossible. An RCT of chloroquine could be done in a week or two and I imagine that is happening now. 30 years perusing the medical literature has taught me that trying to use observational data to make clinical decisions is like chasing squirrels.
Nick:
Good point. Still, even with experimental data, you’ll need some combination of theory and observational data to generalize to real-world populations and conditions.
And sorry, chasing squirrels is a dog metaphor. I should have said chasing butterflies.
Because of people who might die in that “week or two?” Is an RCT superior? Sure. Are there circumstances where you shouldn’t wait for definitive RCT evidence? Yes. Is this one of them? Dunno… depends on the strength of the observational evidence.
RCT for the efficacy signal.
I’d hesitate to privilege safety, tolerability, compliance, adherence, and cost benefit information from an RCT over an obs study.
In situations like this we need interlocking sources of evidence, not a hierarchy.
If your objective is to determine the effectiveness, sure wait for a week or two. With thousands dying everyday and no known effective treatment, a physician shouldn’t wait unless the potential risk of side effect is unacceptable given the patient condition.
I also don’t know if it is ethical to conduct a RCT for critically ill patients that could tolerate the drug. Are such RCTs being conducted or are they mostly conducted on patients with mild to moderate symptoms?
https://web.archive.org/web/20200405061401/https://medium.com/@agaiziunas/covid-19-had-us-all-fooled-but-now-we-might-have-finally-found-its-secret-91182386efcb
Discussion of ER and critical care doctors (“this goes against every dogma”):
https://rebelem.com/rebel-cast-ep79-covid-19-trying-not-to-intubate-early-why-ardsnet-may-be-the-wrong-ventilator-paradigm/
Thanks for these. I know a lot of people get frustrated with Anoneuoid but I think the gadfly approach is important to include in science. This lays out a mechanism that makes sense and at least needs to get tested and some trials directed at these sorts of treatments.
Here is at least some kind of investigation of the biochemistry of this through structural chemical interaction computations:
https://chemrxiv.org/articles/COVID-19_Disease_ORF8_and_Surface_Glycoprotein_Inhibit_Heme_Metabolism_by_Binding_to_Porphyrin/11938173
which is interesting and a bit more useful than some random guy on medium.com who doesn’t cite any references at all.
Anoneuoid: given the biochemical ideas expressed in these papers, what is your opinion on the vitamin C and other antioxidant approaches? Is it primarily keeping tissues from being damaged by peroxides produced by the free radical iron. Does that make sense still? It clearly isn’t going to help with replacing hemoglobin.
Do you think that the smoking link you mentioned in the past could be caused by the fact that smoking damages oxygen transport ability and people’s bodies adapt to produce red blood cells at higher rates? That makes a lot more sense to me than the idea that somehow smoke in your lungs protects your lungs, a theory I find unlikely.
Thanks for your interest Daniel. I started collecting my thoughts and it became a big summary of what is going on afaict.
May as well dump it below:
Clinical stuff
Basically what happened is there were a bunch of large international RCTs done to determine a standard protocol for severe pneumonia and ARDS but the results don’t generalize to this illness (in fact it may be actively harmful, like increasing mortality from 0 to 60% harmful). It took awhile for people to notice because usually the ICU only sees patients after they have already progressed and the ER only sees the initial phases. So specialization was hiding the true picture of the disease.
Now there is a bunch of institutional pushback preventing adoption of a new strategy, which basically amounts to treating the patient instead of religiously following the protocol. Eg in that interview Slidell mentions patients with very low spO2 (% hemoglobin bound to oxygen in the blood) who are clinically not in distress. This normally demands intubation and increasing pressure to push up the spO2, but if it’s a problem with the blood carrying O2 instead of with the lungs this is largely futile. Normally that dangerous treatment is only done because a patient like that is expected to die soon anyway, which doesn’t seem to be the case here.
https://www.medscape.com/viewarticle/928156
https://www.the-hospitalist.org/hospitalist/article/220301/coronavirus-updates/protocol-driven-covid-19-respiratory-therapy-doing
We’ve learned that pretty much all the symptoms mimic high altitude sickness.
Symptoms of COVID-19
https://pubmed.ncbi.nlm.nih.gov/32104915/
https://www.ncbi.nlm.nih.gov/pubmed/32133578
https://www.ncbi.nlm.nih.gov/pubmed/32215618
Symptoms of high altitude sickness
https://www.ncbi.nlm.nih.gov/pubmed/20591371/
https://www.ncbi.nlm.nih.gov/pubmed/29153259
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7096066/
https://www.theguardian.com/books/2007/oct/17/sportandleisure.sport
https://www.ncbi.nlm.nih.gov/pubmed/28947454
Vitamin C
The main thing it does is limits oxidative damage, which is common to every disease process. That’s why it gets made fun of as “magic”. I think they need to measure vitamin c levels and determine if there is a deficiency (as a general rule there will be if they are very sick, exceptions are, eg, if the kidney isn’t excreting anything) then give enough to correct that deficiency.
But, if we assume there is something going on with hemoglobin/iron:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3955505/
However, at high concentrations in the presence of iron ascorbate can also cause a net production of hydroxyl radicals:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5131655/
That could theoretically be a very bad side-effect, but so far the reports out of china do not reflect that:
http://www.drwlc.com/blog/2020/03/18/hospital-treatment-of-serious-and-critical-covid-19-infection-with-high-dose-vitamin-c/
I include the second point just because it also seems to indicate a blood disorder. But if we were going to see a harmful systemic pro-oxidant effect due I’d think it would have happened in the severely ill patient who was determining rapidly. It’s also possible that the peroxyl radical generation is more localized and deactivates the virus… Or maybe all the iron released (assuming it is) is bound to ferritin which is reportedly increased in these patients:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3539674
Smoking
See the above quotes about smoking and high altitude illness. For some reason it seems to help there as well, possibly even having an acute effect in addition to whatever long term effect according to that Captain GJ Finch. Of course, once again, there has been a general refusal by the medical community to figure out what is going on.
One interesting thing I found is that patients with COPD have a lower O2 saturation than usual:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3682248/
Same with asthmatics:
https://www.ncbi.nlm.nih.gov/pubmed/31690381
And also in smokers once you correct for higher carbon monoxide levels:
https://www.ncbi.nlm.nih.gov/pubmed/10070581
So perhaps there is already some adaptation they have to deal with these lower oxygen levels that is also protective against this illness (characterized by low spO2 levels). But I don’t see why that is any more plausible than a remodeling of the respiratory tract (via altering ACE2 expression, or otherwise). Usually when somthing has such a huge effect it is via multiple mechanisms, so I would guess it is both.
Here is a recent meta-analysis on smoking and COVID-19 btw:
https://www.qeios.com/read/article/561
They miss that the same thing was true for the first SARS, a bunch of papers are not included, and that there are no missing smokers for influenza or heart disease, etc. And they say that even if smoking helped with COVID-19 the risks of other diseases would outweigh it (interesting we don’t shut down the world economy over these other diseases). But it is pretty good.
Regarding the ferritin:
https://www.salon.com/2020/04/05/what-it-feels-like-to-survive-covid-19s-dreaded-cytokine-storm/
Ah! I was just about ask if “cytokine storms” were involved, and give the reference https://www.npr.org/sections/health-shots/2020/04/07/828091467/why-some-covid-19-patients-crash-the-bodys-immune-system-might-be-to-blame, that I heard just this morning.
Actually, I assumed the serum ferritin could still bind iron (usually it is intracellular). I can’t find a good source on that.