Moral Anti-Epistemology 2015-04-24T03:30:27.972Z · score: 2 (8 votes)
Arguments Against Speciesism 2013-07-28T18:24:58.354Z · score: 34 (58 votes)


Comment by lukas_gloor on Is Altruism Selfish? · 2020-06-13T19:45:43.027Z · score: 2 (1 votes) · LW · GW

I'm happy to grant you that, when pondering a specific decision, people always choose the option they feel better with in the moment of making the decision. If they have cravings activated, that sense of feeling better will cash out in terms of near-term hedonism (e.g., buying two packs of crisps and a Ben&Jerry's ice cream for dinner). If they make decisions with the brain's long-term-planning module activated, they will make whichever decision they feel most satisfied with as a person (e.g., choosing to do a PhD even though it means years of stress).

No one purposefully makes a decision that predictably makes them feel worse for having made that decision. In that sense, all decisions are made for "self-oriented" reasons. However, that's a trivial truth about the brain's motivational currency, not a philosophical truth about altruism versus selfishness.

Altruism is about taking genuine pride in doing good things for others. That's not what makes altruism "secretly selfish." It's what enables altruism. It also matters to what degree people have a habit of fighting rationalizations and hypocrisy. Just like it feels good to think that you're being virtuous when in reality you're entitled and in the wrong, it also feels good to spot your brain's rationalizations and combat them. Both things feel good, but only one of them contributes to altruistic ideals.

Comment by lukas_gloor on How to learn from a stronger rationalist in daily life? · 2020-05-21T08:58:29.698Z · score: 3 (2 votes) · LW · GW

I recommend finding some kind of goal other goal than "becoming more rational." Going to a workshop here and there or discussing rationality techniques with someone sounds good, but if that's your primary goal for several months or longer, that IMO risks turning into a failure mode of looking at rationality as an end rather than a means. I think you learn most by trying to do things that are important to you.

I strongly agree with the advice of trying to surround yourself with some people you want to learn from.

Comment by lukas_gloor on Why COVID-19 prevention at the margin might be bad for most LWers · 2020-05-17T20:14:13.630Z · score: 2 (1 votes) · LW · GW
We can expect some small regions will make it out with sub 1% but I think there's a 90% chance at least 4% of the US will be antibody positive from exposure (with or without severe symptoms) after a year

That sounds exactly right.

(and a 90% chance no more than 60% will)

I'd say you can go up to 97% for that.

I think the median will be somewhere around 10% of the US population very roughly and that's why I disagreed with the OP. It's unlikely I'd change my mind too drastically about those numbers, at least not in the near future and without new info, because I've spent a lot of time forecasting virus questions. :)

Comment by lukas_gloor on Will the world hit 10 million recorded cases of COVID-19? If so when? · 2020-05-13T20:44:17.998Z · score: 4 (3 votes) · LW · GW

There was a Metaculus question that opened in early April about "How many COVID-19 deaths will be recorded in the month of April, worldwide?" The community prediction was 210k (50% CI: 165k – 288k), which seemed little different from just extrapolating the trend of reported deaths. I saw that countries had all gone into lockdown a while back, so I predicted 75% that the numbers would end up below 193k. The resolution was 184k and I won a lot of points.

Trend extrapolation is only half of what's important. If the trend is foreseeably going to break because circumstances are changing, we need to factor that in. If avturchin is right about the recent numbers being linear with 100K cases a day (I didn't look this up), then we can say that it'll probably take longer than 60 days until 10M confirmed cases. In the majority of locations, R0 is below 1 and many people are recovering (and PCR tests only catch active infections). Of course, case numbers may go up again, which can happen surprisingly fast. Still, I think the mark for 10M confirmed cases is unlikely to be hit before August. Unfortunately, I suspect that we will hit it at some point later in the year when cases go out of control again in some parts of the world where there's extensive testing.

UPDATE June 12th: Seems like I got this one really wrong. Daily new cases are at 135k now, so a substantial increase in cases.

Comment by lukas_gloor on Why COVID-19 prevention at the margin might be bad for most LWers · 2020-05-10T16:44:32.153Z · score: 4 (3 votes) · LW · GW

Thanks for clarifying, that makes sense.

The only strong stance I took (as far as I can see) is that the countermeasures are harmful even without considering their costs.

I think your wording also kind of implied that a large fraction of the population is going to get the virus. Maybe you were primarily thinking of people with jobs that put them at risk, but I think even for those populations, expecting >50% of people with such jobs to get it is very much taking a strong stance. I was wondering if you'd think differently about your dislike of the LW emphasis on advice if you thought that the expert predictions were spot on.

Edit: But maybe that's just not the crux. Maybe you're not saying "you're going to get it sooner or later anyway" but rather "sooner or later, you're going to _decide_ that you're fine with probably getting it anyway."

And that's a stronger argument, I think. But I think a lot of people have probably thought about it, and I don't think keeping your probability of getting this virus <3% is extremely socially restrictive for the rest of how long it'll take. That said, I'm an extreme introvert so probably I don't quite factor in all the things that social people are missing.

Comment by lukas_gloor on Why COVID-19 prevention at the margin might be bad for most LWers · 2020-05-10T14:05:01.797Z · score: 4 (3 votes) · LW · GW

I get the impression that you might be thinking about this in terms of a false dichotomy. It seems correct to me to note that much longer lockdowns are politically infeasible in large parts of the US, but this doesn't mean that most states will just let their entire population catch the virus. Maybe there'll be a second wave and then states that are similarly badly hit as New York and New Jersey will change their stance. Or maybe some states succeed at lessening the restrictions in a smart way, with masks and so on. Maybe the people are sufficiently afraid to catch the virus that they socially distance themselves of their own accord, even when businesses reopen.

Expert predictions say that there have been between 4.8M—28M infections in the US so far (80%) confidence interval. Those infections are responsible for 73k+ deaths so far, and predictions say the median number of deaths will be below 300k in the US (probably even below 200k but I think there was an upward trend in the latest survey, and some chance they're now between 200k and 300k).

I've been doing predicting as well and I agree with those predictions (I'm saying this because it can be justifiable to not always trust experts). Therefore I don't think it makes sense to assume you'll likely be exposed to the virus anyway. (The case for this is even stronger if you live in Germany or the UK; the upward trend in predictions about the US is a bit concerning.) For those who want to avoid low-ish but non-negligible risks of becoming sick for sometimes quite a long time, with a virus that in some instances can do do all kinds of strange and scary things that we don't fully understand yet (see also Elizabeth's comment above), it's good to have the advice available! (Of course, I'm not necessarily saying I endorse all of those pieces of advice.)

Comment by lukas_gloor on April Coronavirus Open Thread · 2020-04-28T12:19:14.878Z · score: 5 (3 votes) · LW · GW

I've heard people with good judgment criticize the Imperial College modelling for countries outside the UK because the forecasts proved to be too pessimistic repeatedly. That's interesting because I know that their UK forecasts were slightly too optimistic. They predicted 20k deaths for the UK initially, then updated to "probably a bit less than that" shortly afterward. And now we're at 21k deaths already (but daily deaths slowed down a lot). I would imagine that their forecasting is the most accurate for the UK numbers because that's what their main task is about.

Comment by lukas_gloor on Kevin's Shortform · 2020-04-25T12:11:09.033Z · score: 4 (2 votes) · LW · GW
What would it take for you to think that it's ok for romantic partners to visit and community house quaranteams to merge?

Ethically, I think this can be fine if strong precautions are taken to avoid infecting non-consenting individuals. (The freedom rhetoric only works if one's actions don't impinge on other people's rights not to be exposed to a deadly illness.) If the only potentially virus-transmitting contacts are with people who follow the same precautions, that's fine in theory. In practice, it can often be difficult to have justified confidence that other people will stick to the rules.

Example: If you infect a person from another household who starts to allow visits with your household under the assumption that both households are otherwise shut off from the outside world, but then one of the people in the other household also makes an exception for visiting her family, and a person from the family gets infected too and goes to grocery stores without a face mask, then you now started a new chain of transmissions that can kill dozens of people who had absolutely no intention of voluntarily taking on additional risks of being infected.

Risking such negative effects may still be justified as long as the probability of it happening is low enough – after all, there are many tradeoffs and we don't prohibit cars just because they foreseeably kill a low number of people. That said, I expect governments to be aware of those tradeoffs. Accordingly, the restrictions should already be lowered soon, and unilaterally lowering them even further can lead to too much tightening up of the network connections between people and households, which could result in an unacceptably high transmission rate. (It's not necessarily just R0 > 1 that's problematic – depending on number of currently active infections, even an R0 < 1 could result in arguably unacceptably many deaths compared to what it would cost to prevent them.)

Comment by lukas_gloor on Kevin's Shortform · 2020-04-25T11:47:02.643Z · score: 4 (2 votes) · LW · GW

>where it's possible that 80% of people have had the virus,

If a demographically representative cross section of the population is infected, I would operate under the assumption that about 0.9% of them will die. From what you write about NY city, it sounds like you think the fatality rate might be a lot lower. I think this will be a major crux for people and so I'd focus first on addressing questions like why recent serology surveys in NYC grocery stroes find only 21% of people with antibodies.

Comment by lukas_gloor on Peter's COVID Consolidated Brief for 2 April · 2020-04-24T00:48:08.091Z · score: 2 (1 votes) · LW · GW

This puts a new light on experts getting the predictions wrong. People are speculating that some of the California cases date back to January or even December. Similar stuff could have happened in New York. IMO, that's the type of thing that makes sense to have outside one's 95% confidence interval.

EDIT: OTOH it seems as though the infections only started in New York in February, and yet they spread to infect a large portion of the population there (tentative serology estimates say about 20% for the city). It doesn't seem to be the case that the wide spread is explained by the infection in New York having started a lot earlier than expected. But something about this confuses me. If the infections reached the Bay area months earlier than they reached in New York, why is New York worse off? I guess one unusually thing about New York is how insanely little space they have inside restaurants and so on. Go to a California Starbucks and it's awesome and comfortable. Go to a New York Starbucks (wasn't it even invented there??) and you can't even sit anywhere and there are walls all around you. Probably infections just spread way faster in that tightly crammed setting?

Comment by lukas_gloor on Jimrandomh's Shortform · 2020-04-15T21:48:01.286Z · score: 7 (5 votes) · LW · GW

What about allegations that a pangolin was involved? Would they have had pangolins in the lab as well or is the evidence about pangolin involvement dubious in the first place?

Edit: Wasn't meant as a joke. My point is why did initial analyses conclude that the SARS-Cov-2 virus is adapted to receptors of animals other than bats, suggesting that it had an intermediary host, quite likely a pangolin. This contradicts the story of "bat researchers kept bat-only virus in a lab and accidentally released it."

Comment by lukas_gloor on Coronavirus: Justified Key Insights Thread · 2020-04-15T18:15:10.712Z · score: 2 (1 votes) · LW · GW

I've looked into this a lot and I agree strongly with this being a good range.

Comment by lukas_gloor on The case for C19 being widespread · 2020-04-13T19:35:16.246Z · score: 4 (2 votes) · LW · GW

Yup. 0.77% is also what I keep stumbling upon when I look into various data points about the IFR! It's my best guess about where Iceland's IFR will end up, and very close to my best guess for proper age adjustment for the Diamond Princess.

Comment by lukas_gloor on The case for C19 being widespread · 2020-04-13T13:26:36.355Z · score: 2 (1 votes) · LW · GW

It's worth noting that the German serology study (it was in the town Gangelt) has been criticized for being poorly presented:

One point of criticism is that the renowned German experts who were asked to comment on the study say they are skeptical about the antibody tests. They argue that to their knowledge, the only antibody tests widely in use in Germany at the time of the study can't distinguish between SARS-CoV-2 and other coronaviruses responsible for a third of common colds. Because we are 1 month past the peak of cold season, they argue that the 15% could be largely picking up on false positives for SARS-CoV-2.

Comment by lukas_gloor on Why I'm Not Vegan · 2020-04-10T11:35:00.664Z · score: 3 (2 votes) · LW · GW

I agree with this if you're comparing complete veganism to something like "reducing one's former consumption of animal products to <10%." But I'd be interested in discussion of the <10% thing. I don't quite like the framing of "purchasing consistency" for that because it doesn't seem like one gets a lot of moral fuzzies from being "sort of almost close to vegan." And many of the arguments against veganism also apply against the <10% thing. And yet, it feels quite problematic to me to think that I don't want to be the type of person who does the <10% thing. What's that driven by? (Not asking you to reply; I'm just thinking out loud.)

Comment by lukas_gloor on Why I'm Not Vegan · 2020-04-09T23:07:22.388Z · score: 6 (4 votes) · LW · GW
This means I'd rather see someone donate $43 to GiveWell's top charities than see 100 people go vegan for a year.

This is saying something different from "I'm not vegan."

I'm not vegan myself either (anymore), but I would care a lot about the impact of 100 people going vegan, and I could imagine so would a lot of non-rationalist meat eaters. Maybe I'm not factoring in how counterintuitive it is how few entire animals are actually eaten by someone, and how effective Givewell charities are by comparison. But on the face of it, this statement feels quite unusual to me.

Edit: I should really have thought about the actual numbers rather than the confounder with money donated to an effective charity. So, according to the post, the comparison is 1 healthy human life year for the following:

  • preventing 80 factory farmed cow years
  • preventing 80 factory farmed pig years
  • preventing 3,300 factory farmed chicken years
  • preventing some % of 300 fish years (representing the %-age of farmed fish rather than wild-caught fish)

I think it's defensible to call this "unusual" but I agree there are many people who would give way higher animal numbers still.

Comment by lukas_gloor on Why I'm Not Vegan · 2020-04-09T13:59:14.132Z · score: 8 (13 votes) · LW · GW
Conditional on animals mattering, how many animal-years on a factory farm do I see as being about as good as giving a human another year of life?

This compares "giving a year of life" to preventing suffering. It's unclear to me whether you're someone who cares unusually little about animals, or whether you're someone who cares unusually much about "giving years of life to self-aware beings that form life plans." Many animal advocates (esp. ones that follow Singer's philosophy) would agree that there's an important difference between human lives and animal lives. But not that there's an important difference about human suffering versus animal suffering.

Comment by lukas_gloor on April Coronavirus Open Thread · 2020-04-08T00:42:21.173Z · score: 8 (4 votes) · LW · GW

If you go through my LW comment history you'll find that I'm in the camp of "The IFR is definitely >0.3%, and very plausibly >0.8%" and that I seem to care somewhat strongly about conveying this to others. :) Maybe you'll find some of the discussions (or links therein) useful. (Unfortunately I can't recommend any single resource that looks super convincing all on its own.)

Edit: By "very plausibly" I mean 25% likely rather than 50% likely. By "definitely" I mean 97% likely.

Comment by lukas_gloor on April Coronavirus Open Thread · 2020-04-08T00:25:19.127Z · score: 5 (3 votes) · LW · GW

Someone in that twitter thread points out that with subtracting false positives, it implies that 10% would be the better guess, as opposed to 13-14%. Does that make sense? Then 4 Covid-confirmed deaths per 620 people would be 0.66%.

And what about sampling bias? I read that the tests were voluntary. Unless someone was extremely meticulous about trying to somehow get a representative sample, I don't think it's reasonable to treat this as random. It's really quite obvious that people who had flu-like symptoms for a couple of days will be more curious to go among people and have a needle stuck into them. .

Comment by lukas_gloor on Would 2009 H1N1 (Swine Flu) ring the alarm bell? · 2020-04-07T17:04:56.747Z · score: 2 (1 votes) · LW · GW
I would also like to investigate this question for MERS, SARS, the 1968 Hong Kong flu, and (as far as it's relevant) the 1918 Spanish flu.

I'd be very interested in analyses of those (esp. if you look at it from the limited perspective people had in the early stages of those outbreaks). I feel like I completely missed it at the time, but the more I hear about SARS-1 the more I feel like the alarm bells should have gone off like crazy (and that probably happened in Asia but the way I remember it, reporting on SARS in the West felt no different from reporting about bird flu or Swine flu – but probably I didn't play close attention because I was really young).

Comment by lukas_gloor on Would 2009 H1N1 (Swine Flu) ring the alarm bell? · 2020-04-07T16:56:42.490Z · score: 4 (2 votes) · LW · GW
>The death rate from swine flu was 0.02%, hitting the young harder than the elderly. I count this as a no.

This is not quite the right way of looking at it! I think you'd have to look into what experts thought during the early months of the Swine flu outbreak. I haven't researched this but I've read that early best estimates for Swine flu fatality were at least a factor of 5 higher than the true infection fatality rate, if not even higher. (The IMO misguded folks who think the IFR for Sars-CoV-2 could be as low as the flu's are constantly pointing this out, failing to flag that this is far from a universal trend among outbreaks – e.g., early estimates of Sars-1 fatality turned out to be underestimates.)

That said, it seems plausible that even with my proposed adjustment, the numbers would still remain below the thresholds you list under "harm." It depends on how much credence experts put on the higher end of the range during the early months of the Swine flu outbreak. I don't know just how high the highest estimates were that still came from credible experts.

Comment by lukas_gloor on Peter's COVID Consolidated Brief for 2 April · 2020-04-07T09:52:08.079Z · score: 2 (1 votes) · LW · GW

I mostly made my comment to point out that the particular question that's being used as evidence for expert incompetence may have been unusually difficult to get right. So I don't want to appear as though I'm confidently claiming that experts need a lesson on forecasting.

That said, I think some people would indeed become a bit better calibrated and we'd see wider confidence intervals from them in the future.

I think the main people who would do well to join Metaculus are people like Ioannidis or the Oxford CEBM people who sling out these unreasonably low IFR estimates. If you're predicting all kinds of things about this virus 24/7 you'll realize eventually that reality is not consistent with "this is at most mildly worse than the flu."

Comment by lukas_gloor on Peter's COVID Consolidated Brief for 2 April · 2020-04-07T00:58:37.848Z · score: 6 (3 votes) · LW · GW

Metaculus (me included) also did similarly poorly on the question of US case growth. Out of all Metaculus questions, this one was probably the one the community did worst on. Technically expert epidemiologists should know better than the hobbyists on Metaculus, but maybe it's a bit unfair to rate expert competence based on that question in isolation.

What was surprising about it was mostly the testing ramp-up. The numbers were dominated by how much NY managed to increase their testing. I managed to overestimate the number of diagnosed cases in the Bay area, while still heavily underestimating the number of total cases in the US.

This is the relevant Metaculus question:

If you look at the community median at a similar date to the prediction by expert epidemiologists, it's also off by a factor of 6 or so. (Not sure what the confidence intervals were, but most likely most people got negative points from early predictions.)

(For those interested, the Metaculus user "Jotto" collected more examples to compare Metaculus to expert forecasters. I think he might write a post about it or at least share thoughts in a Gdoc with people who would be interested.)

Comment by lukas_gloor on What is going on in Singapore and the Philippines? · 2020-04-06T13:15:32.678Z · score: 13 (8 votes) · LW · GW
  • Heat and humidity probably slow down the transmission rate, but not enough to make large outbreaks impossible.
  • I could imagine that heat and humidity are especially beneficial for countries during the containment phase (esp. for contact tracing). According to this interview, the virus is inactivated at temperatures 30 degrees or higher. This could reduce the number of transmissions in setting that are particularly hard to contact trace (public transport, small grocery stores). As long as transmissions happen primarily in air-conditioned buildings or household contexts, contact tracing is much easier. (But perhaps it was doomed from the start, and the heat/humidity only meant it took longer to notice the cases that were being missed.)
  • Singapore and the Philippines seem very different to me in several respects!
  • The Philippines had reported 8 deaths by March 15th already. That's indicative of a large undetected outbreak early on. I know almost nothing about how much testing they've done, but I could imagine that it's not a lot. I could imagine that deaths in the Philippines are vastly underreported even now.
  • By contrast, Singapore definitely seemed to have their outbreak under control initially. I think there's a good chance it could have worked with earlier border closures. They only closed borders on March 24th after several imported cases, primarily from Indonesia.
  • Indonesia (which also has had hot climate throughout recent events) has one of the highest deaths-to-confirmed-cases ratios worldwide, and that's not factoring in that they may have missed >1,000 deaths already. According to that Reuters article, Indonesia had conducted only about 7,500 tests by April 3rd. By comparison, the UK conducted more tests on April 3rd itself (in a single day) even though its population is 3x lower than Indonesia's. Experts had been saying all along that Indonesia not reporting any cases throughout February was extremely suspicious based on travel connections to Wuhan. It seems that they were spot on. I think it's quite likely that Indonesia has >100,000 active cases by now. This suggests to me that dozens of Indonesians must have imported the virus to Singapore before the border closure (though maybe they all underwent temperature checks at the very least, and possibly quarantine?).
  • An alternative hypothesis (or contributing factor at the very least) is that containment failed because Singapore did not recommend mask usage as much as Hong Kong for instance did. Probably that was partly because of limited supplies, though the way it was communicated was similar to CDC communications ("masks don't help unless you're sick"). It seems increasingly likely to me that outbreaks are very hard to contain without widespread usage of masks (South Korea and Hong Kong rely heavily on mask usage – maybe someone could check up on the situation in Taiwan to get more data points on this).
Comment by lukas_gloor on Atari early · 2020-04-02T18:36:25.955Z · score: 5 (3 votes) · LW · GW

Another thing is that the bots never make exploits. So when there's a bad player at the table playing 95% of their hands, the bot would never try to capitalize on that, whereas any human professional player would be able to make extra money off the bad player. Therefore, the bot's advantages over human professionals are highest if the competition is especially though.

Comment by lukas_gloor on Atari early · 2020-04-02T08:30:57.749Z · score: 8 (4 votes) · LW · GW
I’m not familiar enough with Poker to say whether any of the differences between Texas Hold’em, Omaha Hold’em and Seven Card Stud should make the latter two difficult if the first is now feasible.

I've played all of these and my sense is that Seven Card Stud would be relatively easy for computers to learn because it has fixed bet sizings just like Limit Holdem, which was solved long before No Limit. Some of the cards are exposed in Stud which creates a new dynamic, but I don't think that should be difficult for computers to reason about that.

Omaha seems like it would be about as difficult as Texas holdem. It has the same sequence of actions and the same concepts. The bet sizings are more restricted (the maximum is determined by the size of the pot instead of no limit), but there are more cards.

As far as I'm aware, none of the top poker bots so far were built in a way that they could learn other variants of poker without requiring a lot of fine-tuning from humans. It's interesting to think about whether building a generalized poker bot would be easier or harder than building the generalized Atari bot. I'm not sure I know enough about Atari games to have good intuitions about that. But my guess is that if it works for Atari games, it should also work for poker. The existing poker bots already rely on self-play to become good.

Comment by lukas_gloor on Coronavirus: California case growth · 2020-04-01T13:31:31.963Z · score: 5 (3 votes) · LW · GW
It seems like you downvoted because you think I used a serious tone when the point I wanted to make was minor. I think you made a mistake and assessed the situation wrongly.

Yes, this is what happened. I didn't read closely enough and I thought what Vipul decided to call "true cases" was simply the total number of infections. But he wanted to specifically refer to only the infections that were going to become symptomatic at some point. I agree that this is making a distinction that doesn't carve reality at its joints. On top of that the label seems to have misleading connotations (evidenced by me having misunderstood what he meant:)). I agree that this can be risky in this context especially.

I'm reversing the downvote! I don't see though how outsiders could have immediately inferred from your comment that you object to how Vipul drew categories instead of merely his use of non-standard terminology. I think it's innocuous to use non-standard terminology if one is not the WHO, and if the choice of terminology is intuitive and carves reality at its joints.

And about the WHO example, I totally agree. I criticized the WHO for the same reason here:

Comment by lukas_gloor on Coronavirus: California case growth · 2020-03-31T20:21:38.055Z · score: 4 (2 votes) · LW · GW
the tone of a comment shouldn't be as an important consideration as the point being made.

Tentatively agree, but in this case the point was about a mostly aesthetic (though common) preference for established terminology, which has nothing to do with anything of substance. It's fair to point out that not everyone cares equally little about written appearances, but it seems uncalled for to frame it in a way as though the author was violating a norm. (If people now want strict academic norms for a community blog that initially was started by Eliezer Yudkowsky of all people, that's another discussion.)

I still think rationality means thinking rationally.

One feature of that is to notice instances when usually sound heuristics are drifting apart from the actual goal. Some people can't help but feel increasingly more averse to making posts on here if they frequently encounter feedback that makes them feel as though they did something wrong for sharing their thoughts in a suboptimal fashion. Maybe you're not high on neuroticism, maybe ChristianKI isn't high on it, and maybe Vipul isn't either. But I wouldn't be surprised if people high on neuroticism are overrepresented among rationalists – maybe just not among the ones who frequently post here (and that's my point). So just because some people wouldn't get discouraged by slightly pedantic criticism worded in a judgmental fashion doesn't mean it's not discouraging for anyone. And it doesn't help that you're implicitly suggesting that people are being less rational if criticism affects them more. If some portion of the population is afraid of spiders, you don't throw spiders at them and say "being rational is about not being affected by negative emotions." Okay, bad analogy: Criticism usually correlates with truth seeking; throwing spiders does not. However, I think many of the people who are unusually discouraged by judgmentally-worded criticism are discouraged precisely because they take criticism in general unusually seriously. That's often a virtue. I think LW culture has drifted toward an equilibrium where some traits that usually correlate with rationality are rewarded too much, and other qualities, which can often be virtuous too (in the right person/combination) are written off as attempts to undermine truth seeking. I think that's a an example of a common failure mode for communities, where signalling dynamics combine with selection effects created by the signalling until what's left is a culture that is unhealthily extreme on some dimensions, but few in that culture are aware to notice.

(And I've a couple of big doses of unexplained negative karma on the posts I've created and would have much preferred some comment/feedback whatever the tone it took and Christian was one of the few that provided some.)

It's good when people explain why they downvoted something, and I think harsh feedback can be really valuable. I also realize that for some people it's difficult to word their feedback nicely (this applies to me too if it concerns a dimension I strongly care about). Usually I agree with your sentiment that it's better to get the criticism in whatever form, if the alternative is not hearing it at all. But that stops to apply if the points are sufficiently minor and the tone sufficiently discouraging. (And continuing to try to give feedback well continues to be important even if we – reluctantly rather than triumphantly – have to agree that lapses are usually to be excused for the greater good of rationality.

Comment by lukas_gloor on Coronavirus: California case growth · 2020-03-31T02:29:02.412Z · score: 4 (2 votes) · LW · GW
I don't think this is a time to make up new LW terminology without good reason. It would be worthwhile to look up the established term from the literature before making up terms like this.

Downvoted for tone and and the effect I tentatively think this might have on people's motivation to go through the trouble of writing up their interesting ideas.

(If you want LW to become increasingly more similar to a forum for academic discussions, then sure, might be good to give feedback this way. But I don't see why that should be the primary aim.)

Comment by lukas_gloor on Iceland's COVID-19 random sampling results: C19 similar to Influenza · 2020-03-29T01:53:49.781Z · score: 8 (2 votes) · LW · GW
In terms of true number of infected, I'm predicting that SK has on the order of 100K to 200K cases and say 4K in Iceland, and I don't find this up to ~50x difference very surprising. Firstly, it's only about an 18 day difference in terms of first seed case at 25% daily growth.

I see. BTW our confidence intervals for the IFR have some overlap: 0.4% is my lower bound and your higher bound. :)

Comment by lukas_gloor on Iceland's COVID-19 random sampling results: C19 similar to Influenza · 2020-03-29T00:35:05.216Z · score: 12 (4 votes) · LW · GW

Edit: changed some numbers slightly after looking things up, to make them more accurate.

Modeling one country as "X weeks behind" some other country is hazardous at best and also unnecessary as Iceland provides direct graphs on their daily #tests and #positive.

I agree that it's tricky to do the modelling correctly, but I feel like you're not engaging with my point properly. I think the following argument I made is watertight:

  • There was a point when South Korea had several deaths (50ish) and thousands of cases (7,700) and their IFR was at 0.6%.
  • That's roughly when they got their outbreak under control. The numbers slowed down tremendously, and 20 days later they are only at 9,500.
  • So in those 20 days, the reported CFR 2.5xed.
  • Iceland's reported CFR never 2.5xed so far.
  • Therefore, they are way behind South Korea's timeline even if we grant the point that Iceland has its outbreak contained (you may be completely right about this, because I didn't follow it EDIT: I don't think you're right about it because Iceland's numbers grew by almost 10% two days ago, which is still a somewhat large portion of new cases!).

The way I see it, this point is only wrong if somehow Iceland going from 1 deaths to 2 deaths is the equivalent stage of the timeline as South Korea's deaths going from 50 to 144 (or whatever the numbers were). That seems highly improbable to me because it would mean that South Korea's outbreak was 50 times larger than Iceland's. That doesn't seem right to me. (Though I guess if I had a strong belief that the hypothesis you're defending is consistent with other data points, then this may not be a knockdown argument by itself? Would you expect South Korea's outbreak to be 50x larger? No need to answer, of course. But if this argument updates you somehow, I'd be curious to hear!)

Comment by lukas_gloor on Iceland's COVID-19 random sampling results: C19 similar to Influenza · 2020-03-29T00:12:52.268Z · score: 5 (3 votes) · LW · GW

This person has been collecting reports in Italian media and also contact the mayors of Italian district to request the information.

It's not for all of Northern Italy, but it's also not villages either. Those cities or provinces are much more populated than I initially thought (see Stefan Schubert's correction to my comment elsewhere).

Comment by lukas_gloor on Iceland's COVID-19 random sampling results: C19 similar to Influenza · 2020-03-29T00:05:20.430Z · score: 7 (4 votes) · LW · GW
Additionally, if most cases were asymptomatic or weakly symptomatic there would be few cases of multiple close contacts becoming ill. These are common.


I think this point is really underappreciated.

Comment by lukas_gloor on Iceland's COVID-19 random sampling results: C19 similar to Influenza · 2020-03-29T00:02:38.503Z · score: 3 (2 votes) · LW · GW

Minor point because I agree with all the other things you said: While it's true that South Korea didn't have a China-style lockdown, I think the behavioral changes at the city level must have been really quite extreme. Perhaps in part also culture-driven rather than government imposed (maybe South Koreans actually followed government recommendations almost perfectly?), but I think it could be overupdating on the evidence to assume that South Korea didn't need some kind of "lockdown" (loosely spoken) to get the situation under control initially. I'm not 100% sure this is what happened, but I heard at least one expert say that people who claim South Korea didn't have a lockdown are being misleading, and that point also seems to make sense based on priors.

Comment by lukas_gloor on Iceland's COVID-19 random sampling results: C19 similar to Influenza · 2020-03-28T23:50:12.470Z · score: 10 (3 votes) · LW · GW
So the testing of several hundred thousand cult members pushed both their CFR and test positive fraction lower than it otherwise would be, and rather obviously skewed their case age structure.

It skewed the age structure toward a younger demographic. Were you aware of this or did you assume that the religious group is skewed toward old people like typical churches? I didn't realize this up until like ten days ago, but the Christian cult was predominantly pretty young people!

Nonetheless they have tested far less of their population than Iceland (about 5X less as of 3/20 according to ourworldindata), so if the ratio of infections/cases is 4x to 5x in Iceland it seems reasonable that it's 10x to 20x in SK.

The reason I don't consider it at all plausible that South Korea missed 80% or more of its cases is because of how quickly and lastingly they were able to gain control over their outbreak.

And about Iceland: Isn't it really very clear that Iceland is weeks behind South Korea, and that Iceland's numbers are therefore unrepresentatively low? For comparison, South Korea's IFR was 0.6% at a point when they had 7,700 confirmed cases. I think this was roughly 20 days ago. So 20 days for South Korea's IFR to go from 0.6% to 1.5% is how long it takes a majority of patients to die if the hospital conditions are favorable enough to give everyone good treatment. There weren't many new confirmed cases in the meantime because the current count is 9,500. So with respect to the IFR Iceland is currently at (0.21%), if Iceland had their outbreak under control, we should expect that IFR to rise by a factor >2.5x. 2.5x is the lower bound because South Korea's IFR was at 0.6% at a time when they already had dozens of deaths; by contrast, Iceland only has two deaths so they are way behind the timeline. (This comes from the effect that once true cases stop growing, the CFR rises up until all the illnesses take their course.) Expecting anything lower than a 4x increase from time delay is unreasonably low. So to make Iceland's reported CFR comparable to South Korea's, we should think of it as 0.84% rather than 0.21%. And then we can think about how many cases went undiagnosed in both countries (but maybe you did factor this in).

In addition, we have to factor in that Iceland doesn't have their outbreak under control. Or do they? I didn't check up on this, but I'd be surprised if they had the outbreak contained. My guess is they caught fewer cases than South Korea! Yes, Iceland did more testing per capita. But South Korea knew where to look! They really managed to get their outbreak under control. It's very impressive and I feel like they're not getting the credit they deserve.

Anyway, assuming Iceland still has community transmission, this would mean that through new testing, new confirmed cases will be added constantly to the total. Those cases will predominantly be recently confirmed cases where not enough time had passed for people to die. This will keep Iceland's reported CFR at a low level for quite a while to come, but this provides basically zero evidence for the actual IFR being low.

UPDATE: I ended up looking up Iceland's numbers, and it seems like they had almost 10% of their total cases confirmed only yesterday. So whether the growth regime is "linear" or not, I think this is definitely not comparable to South Korea's numbers where the growth has been around 1% or 1.5% for several weeks.

Comment by lukas_gloor on The case for C19 being widespread · 2020-03-28T23:05:14.235Z · score: 9 (4 votes) · LW · GW

@Hauke Hillebrandt

FWIW, if UK death toll will surpass 10,000, then this wouldn't fit very well with this hypothesis here.

If this update works then I feel like just looking at how the numbers in Italy came together would change your mind about the low-IFR hypothesis.

Alternatively, if the Covid-19 deaths in NY state go above 3,333 in the first week of April, that seems like it would also falsify the hypothesis. (NY state has fewer than one third the population of the UK.) Unfortunately I think this is >80% to happen.

Comment by lukas_gloor on The case for C19 being widespread · 2020-03-28T22:18:39.359Z · score: 17 (6 votes) · LW · GW
The "COVID-19 is similar to influenza" model predicts IFR in the 1% range for a retiree age distribution like on DP but 0.1% range on the US age distribution.

FWIW I tried to do an age adjustment for the Diamond Princess myself and what I got was that the 1.4% IFR for the cruise demographics translates into a 0.3% IFR for US demographics (factoring out gender adjustments). I think you could argue that because women were underrepresented on the cruise ship, the adjustment should be greater, so 0.25% is plausible. That said, this doesn't yet factor in that the people who are medically worst off probably don't book cruises, so my best-estimate adjustment is maybe 0.4% with a lot of uncertainty. I agree that the people who use the Diamond Princess as evidence for an IFR around 0.9% or higher seem to be making a mistake. At the same time, I do think the Diamond Princess is at least weak to moderate evidence against the 0.125% figure Ioannidis arrived at, or the 0.1% figure that I've seen discussed elsewhere.

I don't really know how this compares to flu mortality, but I found myself somewhat skeptical about the claim I quoted above. You seem to get a 10x update for your age adjustment, whereas my update was only about 5.7x (before factoring in harder-to-quantify assumptions that IMO reduce the factor a bit more even).

(I made a huge mess of my calculations and I don't recommend clicking on the following link, but just so people see that I'm not just making this up, here's some evidence that I did something with numbers. Could also be that I neglected some considerations. For factoring in how much overrepresentation of age bracket 70-79 changes things, I based the adjustment off of previous estimates on how strongly Covid19's IFR is age skewed. I'd imagine that this adjustment was uncontroversial because whether you subscribe to the low IFR theory or not, probably there's no reason to question whether the proportionalities of the attack rate are correctly reported?)

Comment by lukas_gloor on The case for C19 being widespread · 2020-03-28T11:29:06.858Z · score: 6 (4 votes) · LW · GW
This is a non-random village in Italy, so of course, some villages in Italy will show very high mortality just by chance.

It's extremely implausible that it would be 10x or 15x higher than what's expected for the typical Italian village. Besides, other villages like Cremona or Bergamo also seem to be close to those numbers. Smoking or age structure or air pollution doesn't give you a 10x update.

UPDATE: Wow, I was totally wrong about those being villages. As Stefan Schubert pointed out, those are cities and provinces with tens and hundreds of thousands of inhabitants!

Comment by lukas_gloor on The case for C19 being widespread · 2020-03-28T10:10:35.146Z · score: 3 (2 votes) · LW · GW
In Italy, with almost 10k deaths it would be 0.02%-0.04%

There's an Italian village where 0.1% of the population already died with a confirmed diagnosis of Covid-19. Inferring from typical monthly death rates it's also estimated that the twice as many people died from Covid-19 in that village without an official diagnosis. There's a bunch of uncertainty about those additional 0.2%, but it would put the fatality rate at 0.3% already. And those figures are from 4 days ago (edit: 6 days ago actually).

Edit: It's a province and city(!), not a village.

Comment by lukas_gloor on What is the typical course of COVID-19? What are the variants? · 2020-03-26T00:10:14.772Z · score: 1 (1 votes) · LW · GW
One interesting thing is that people in their 20s and 30s had a much higher rate of symptoms (80%+) than older or younger people (< 60%).

That's indeed interesting. This article seems to say that it's different in young people who tested positive in South Korea.

One point of criticism about the link included under "Diamond Princess data:"

The abstract reads as follows:

Comparing deaths onboard with expected deaths based on naive CFR estimates using China data, we estimate IFR and CFR in China to be 0.5% (95% CI: 0.2–1.2%) and 1.1% (95% CI: 0.3–2.4%) respectively.

This wording to me suggests that the authors think 0.5% is most likely the correct IFR for China's cases (up to a certain point in time, based on when the paper was written). This is either false, or the authors are making a really obvious mistake. The paper did not factor in that sick people in Wuhan (where most of China's cases were from at the time) probably had much worse treatment prospects than patients from the cruise ship.

Comment by lukas_gloor on What is the typical course of COVID-19? What are the variants? · 2020-03-26T00:01:03.661Z · score: 11 (4 votes) · LW · GW

Excellent analysis! I've used the 18% point estimate from that Diamond Princess study without noticing that the math could be off.

One point to add: I've seen people say that asymptomatic presentations of SARS-CoV-2 infection might more common in young people, especially in the age range from 20-40. That age range was underrepresented on the cruise ship. For that reason, perhaps it's possible for up to 65% of cases to be asymptomatic?

That said, I very much agree with you that the entire thing about asymptomatic presentations could be a myth based on false positives and confusing "asymptomatic" with "pre-symptomatic." This study is the type of thing that would give us confidence in the existence of asymptomatic carriers – if only it had more examples than just one person.

Comment by lukas_gloor on Preprint says R0=~5 (!) / infection fatality ratio=~0.1%. Thoughts? · 2020-03-24T14:15:53.240Z · score: 1 (1 votes) · LW · GW

I read about the new deaths on the Wikipedia article.

A Canadian man in his 70s died on 19 March, making him the ninth coronavirus-related death from the ship.[102][46] Two Japanese passengers in their 70s died on 22 March.[47]


I know that. If you follow this discussion up to the beginning, you'll see that all I'm claiming is that the number of documented cases has been affected by selective bias, because asymptomatic / pre-symptomatic etc. cases are unlikely to be diagnosed.

Okay. I feel like the discussion is sometimes a bit weird because the claim that there are a lot of undocumented cases is something that both sides (high IFR or low IFR) agree on. The question is how large that portion is. You're right to point to some sampling biases and so on, but the article under discussion estimates an IFR that it at least a factor 5 below that of other studies, and a factor of 4 (or 3.5 respectively) below what I think are defensible lower bounds based on analysis of South Korea or the cruise ship. I don't think selection bias can explain this (at least not on the cruise ship; I agree that the hypothesis works for China's numbers but my point is that it conflicts with other things we know). (And I already tried to adjust for selection bias with my personal lower bounds.)

I'm not saying this study is right. I'm just saying that, unless someone points a methodological flaw, "their conclusion is too different" is not a reason to discard it.

It depends on the reasoning. We have three data sets (there are more, but those three are the ones I'm most familiar with):

  • South Korea
  • The Diamond Princess
  • China

How much to count evidence from each data set depends on how much model uncertainty we have about the processes that generated the data, how fine-grained the reporting has been, and how large the sample sizes are. China is good on sample size but poor in every other respect. The cruise ship is poor on sample size but great in every other respect. South Korea is good in every respect.

If I get lower bounds of 0.4% and 0.35% from the first two examples, and someone writes a new paper on China (where model uncertainty is by far highest) and gets a conclusion that is 16x lower than some other reputable previous estimates (where BTW no one has pointed out a methodological flaw either so far), it doesn't matter whether I can find a flaw in the study design or not. The conclusion is too implausible compared to the paucity of the data set that it's from. It surely counts as some evidence and I'm inclined to move a bit closer to my lower bounds, all else equal, but for me it's not enough to overthrow other things that I believe we already know.

Comment by lukas_gloor on Preprint says R0=~5 (!) / infection fatality ratio=~0.1%. Thoughts? · 2020-03-24T01:32:15.421Z · score: 1 (1 votes) · LW · GW

They write "at the time of testing." The study I cite followed up with what happened to patients.

Also relevant: In the last 5 days, 3 more people who had tested positive on the Diamond Princess died. And one person died two weeks ago but somehow it wasn't reported for a while. So while my own estimates were based on the assumption that 7 / 700 people died, it's now 11 / 700.

Comment by lukas_gloor on March Coronavirus Open Thread · 2020-03-21T17:17:48.536Z · score: 1 (1 votes) · LW · GW
Older people are more likely to get infected, so the infected population in the US will lean older as well--closer to the distribution on the ship.

Interesting! Do you think this is established? I haven't looked into this, but my guess would have been that the risk is similar because young people are less scared of the virus. But yeah, good point about further adjustments being needed to get the best estimate.

Comment by lukas_gloor on Preprint says R0=~5 (!) / infection fatality ratio=~0.1%. Thoughts? · 2020-03-21T00:05:05.237Z · score: 5 (4 votes) · LW · GW

From the paper:

With testing capacities of 20,000 tests daily, it [South Korea] has the largest and most accurate coverage compared to all other countries as of writing. The low false-negative rate in detecting COVID-19 infections leads to the lowest death rate compared to all other countries (0.84) with major case count

Note that South Korea's reported (naive) CFR is at >1% by now. It's possible that the authors adjusted for the fact that most of South Korea's cases were still active at the time of writing (about 55-60% of cases are still active now, I think), but I don't see this in this paper. It probably doesn't make a huge difference, but still relevant that this could cause the estimates to be a bit too low.

Comment by lukas_gloor on Preprint says R0=~5 (!) / infection fatality ratio=~0.1%. Thoughts? · 2020-03-20T23:59:00.507Z · score: 5 (3 votes) · LW · GW

From the paper:

This method requires the comparison of two countries with sufficient confirmed cases and reported deaths. One country (target country) will be adjusted, given the information from the second country (benchmark country). In order to adjust for the difference in the population demographics of the target country, T, and the benchmark country, B, we compute a Vulnerability Factor (VTB).

Am I right that they're not factoring in that patients had worse prospects in Wuhan than in South Korea? I feel like whatever the outcome of their adjustment process, that value would need to be multiplied by a factor >1 which represents hospital overstrain in Hubei, where at least 60% of China's numbers stem from (probably more but I haven't looked it up). I don't know how large that adjustment should be exactly, but I find it weird that there's no discussion of this. Am missing something about the methodology (maybe it factors in such differences automatically somehow)?

Ah, OK: They list this as an assumption:

[Assumption]Treatment has minor influence on outcome The provided healthcare in countries is comparable. For developed countries such as Italy and South Korea, it is assumed that the population has similar access to treatment. The death rates reported by age group are thus applicable in all countries

This is important to keep in mind when we try to derive implications from their estimate. Especially if we look at the hospitalization rates estimated here on page 5. For this disease in particular where people sometimes have to stay in hospitals for several weeks, it's hard to imagine that treatment only makes a small difference.

Comment by lukas_gloor on Preprint says R0=~5 (!) / infection fatality ratio=~0.1%. Thoughts? · 2020-03-20T22:45:15.771Z · score: 1 (2 votes) · LW · GW

More points in favor of a higher IFR:

  • The percentage of asymptomatic cases on the Diamond Princess was even lower than 50%. It was only about 18%. (I trust this figure because the paper has author overlap with the paper that gave a higher figure initially, and it's written by the same author who made the 0.1% estimate and we'd expect this person to – if anything – have a bias toward expecting a larger number of asymptomatic cases).
  • About the age distribution on the Diamond Princess: I tried doing age adjustment for it here. ((Edited because I revised some estimates.))
Comment by lukas_gloor on March Coronavirus Open Thread · 2020-03-20T20:19:34.262Z · score: 9 (3 votes) · LW · GW

I also thought that in Lombardia, the estimates given by Ioannidis are rapidly trending toward coming in contradiction with SIR models. :( Lombardia has a population of 11 million people and 2,500 reported deaths. Some doctors are raising alarm that many deaths are going undetected because people are dying at a rate that's 4 times higher than the same month last year. In addition, the death counts always lags behind because some people are sick for a long time before they die (though maybe this start to be the case less strongly in conditions of extreme hospital overstrain). All of this suggests that an estimate of 10,000 deaths for Lombardia alone might soon prove to be accurate. But according to the IFR provided by Ioannidis, this would correspond to an expected 8 million people infected (72% of the population). I don't understand SIR models well enough to calculate what the R0 would have to be for 72% of a population to get infected. I suspect that Covid-19's R0 is high enough to be consistent with this, but it wouldn't leave a lot of room for estimation errors.

That said, I think the above calculation is naive, so the argument doesn't work (at least not in this crude form). If hospitals become as overwhelmed as they are in Italy, I'm sure that even someone with Ioannidis' view would expect the IFR for Lombardy to become a lot higher than 0.125% because a lot of people aren't getting life-saving hospital attention.

So, this means that Lombardy isn't necessarily a knockdown argument against Ioannidis's estimate in the same way South Korea is. However, I think Ioannidis's estimate would have counterintuitive implications for the percentage of people infected in Lombardy. It would have to be in the double digits already at the very least. The most trustworthy estimate I saw about Wuhan suggested that only 5% of its population had the virus. However, there's some disagreement about this, and the people who tend to argue for an unusually low IFR also tend to argue that there's a giant iceberg of undetected asymptomatic cases.

UPDATE: I just realized something: I read somewhere recently that Italy is doing 30,000 tests a day by now, and that about 25% of them are positive. This seems to be in contradiction with Ioannidis's estimate because his view should imply that, if there's some kind of selection at all for who they are testing (as opposed to just testing members of the population at random), then we should expect to see more positive test results than 25%. (Why? Because if we assume that hospital overstrain increases his death rate estimate by a factor of 7x (which is a really large adjustment!), the death count estimates for Lombardy combined with Ioannidis's estimates would still suggest that above 10% of the population would have the virus. Such high numbers would only be consistent with reality if most people had relatively mild symptoms or no symptoms at all, so assuming that there's substantial pre-selection on who is getting tested (as opposed to random testing, which would be odd), a rate of 25% positive tests would be implausibly low for the scenario where >10% of the region were infected. So, to conclude, I think one can plausibly construct a case against Ioannidis's estimates based solely on common sense and numbers from Lombardia. I probably haven't quite succeeded at making this case in a watertight way, but I think you might be right with your intuition. This is just one more reason why the 0.125% estimate is completely absurd.

Comment by lukas_gloor on March Coronavirus Open Thread · 2020-03-20T19:43:59.598Z · score: 4 (2 votes) · LW · GW

Reading the Ioannidis article, it seems to say that he did his own calculations, and he doesn't show them. Okay.

I'm curious about this, so I'm going to try a ballparking estimate myself.

Tl;dr I intially arrived at a result that suggested 0.125% was way off, but then found better info on the cruise ship's age distribution and had to revise my judgment. I now find it debatable whether 0.125% is defensible or not, but it's not "way off." My own estimate would be more in the ballpark of 0.3%, but I don't anymore consider the cruise ship to be evidence for IFR estimates at 0.5% or higher.

Update March 24th: In the couple of days, 3 new patients who had tested positive on the Diamond Princess have died. In addition, the Wikipedia article has been edited to list another death that previously hadn't been included. So total deaths per confirmed cases on the Diamond Princess are now 11 / 700 instead of 7 / 700. All my calculations below are based on the older, outdated numbers. To get the most updated estimates, just multiply the results below by 11/7.


Note that I have never done age adjustments for anything, so I have no idea what the proper methdology would be. I'm just curious to see if 0.125% is potentially reasonable rather than (as my current intuition suggests) very dubious.

From this paper, I found the following info:

A total of 634 people tested positive among 3,063 tests as at 20 February 2020. Of 634 cases, a total of 313 cases were female and six were aged 0–19 years, 152 were aged 20–59 years and 476 were 60 years and older.

At the end of the outbreak, roughly 700 people had tested positive. I'm going to assume that the 66 patients not yet in the above statistics fall into age categories in the same proportion. So a bit more than two thirds of the 66 patients get added to the 476 figure for people aged 60 and older.

With this adjustment, we have 700 diagnosed cases, of which an estimated 525 patients were aged 60 and older. Of those 700 diagnosed cases, 7 people died. 525 out of 700 corresponds to 75%. (I'm going to mostly ignore the death risk for people below age 60 for the analysis below, because it will be negligible given that people older than that anyway make up the majority share.)

This wikipedia article on US demographics says the following:

  • 0–14 years: 18.62%
  • 15–24 years: 13.12%
  • 25–54 years: 39.29%
  • 55–64 years: 12.94%
  • 65 years and over: 16.03%

Eyeballing this, let's go with 22% of the population at age 60 or older.

75 divided by 22 is roughly 3.4, so this naively suggests that the cruise ship's demographic was roughly 3.4 times more susceptible to dying from SARS-CoV-2. If I divide the observed IFR of 1% by 3.4, I get 0.3%. Why does Ioannidis get 0.125% instead of 0.3?

Moreover, it seems to me that 0.3% must be an underestimate because I assume that even though the cruise ship population is substantially older on average than the US population, I would think that this effect will disappear (or even reverse) at the extremes, once we look at the percentage of exceptionally old people (e.g., aged 80 and above, age 85 and above, etc.). Because Covid-19 is particularly fatal for the very oldest people, I expect the 0.3% figure to contain a substantial degree of overcorrection. Especially also because elderly people with the most severe pre-existing health conditions are likely heavily underrepresented on cruise ships. This effect could be really quite significant: It's not even totally obvious that a downward adjustment of the 1% IFR observed on the Diamond Princess is warranted. It's probably warranted, but depending on how strongly cruise ship passengers are pre-selected against having unusually bad health, and depending on how strongly pre-existing health conditions affect someone's survival prospect for Covid-19, it's conceivable that the 1% figure doesn't need to be downward adjusted at all.

To conclude, I don't understand how age adjustments for SARS-Cov-2 infections on the Diamond Princess can drive down the estimated IFR substantially below 0.5%. 0.5% seems closer to a lower bound to me than anything else. (Of course, those are point estimates. I don't have strong views on whether 0.125% is outside some appropriate confidence interval, but my impression was that 0.125% was Ioannidis's point estimate, and interpreted as such, it seems clearly much too low!)

UPDATE: Oh I see. I found an age table that I had overlooked initially. It turns out cruises are really popular for people at age 70-79 (there are about 20% more people of that age than 60-69, whereas it's the other way around for US demographics). This distribution makes Ioannidis's figures look more plausible, though the difference doesn't seem large enough to fully bridge the gap between 0.3% and 0.125%, especially because the 80-89 bracket seems to be represented proportionally again. Still, I don't anymore think that 0.125% is horribly off.

Comment by lukas_gloor on Preprint says R0=~5 (!) / infection fatality ratio=~0.1%. Thoughts? · 2020-03-20T17:02:31.132Z · score: 11 (3 votes) · LW · GW
Also this seemingly squares more with John Ioannidis take on Corona:

Ioannidis makes this claim:

Projecting the Diamond Princess mortality rate onto the age structure of the U.S. population, the death rate among people infected with Covid-19 would be 0.125%.

I don't find a source for this. The adjustments I saw looked different. If he's right about those 0.125%, that would be an important update!

But it feels more plausible to me that the 0.125% thing went wrong somewhere because it just seems ruled out by South Korea, which unlike European countries has their outbreak contained. I can't see how South Korea could somehow have missed 700% of their reported cases even though they are conducting 10,000 tests daily, and have fewer than 10,000 confirmed cases.

UPDATE: I took a shot at doing the age adjustment myself here. The summary: I don't see how one can get anything below 0.3% and, adjusting for selection effects where the least healthy people probably avoid going on cruises, even going below 0.5% seems implausible to me. UPDATE2: I adjusted my estimates after finding more precise data. I still think 0.125% is too low, but I think something like 0.2% is perhaps already defensible. This suggests that the estimate was closer than I thought and I now consider the Diamond Princess not to be evidence in favor of IFR of 0.5% or higher (assuming no hospital overstrain).