@(kokotajlod) **Data points for countries that frequently were above 30 degrees Celsius:** Singapore: 100+ confirmed cases, some community spread but it was successfully contained. Very well-prepared for outbreak detection and containment. Virus spread may have happened indoors under strong air-conditioning. Thailand: 40+ initial cases. Very well-prepared for outbreak detection and containment. Thailand ranked first for the most at-risk countries in a report from late January, and yet they didn't have sizeable community transmission. Indonesia: On...

One reason in favor of canceling that I haven't seen discussed yet: I think it'd be difficult to get attendees to talk about stuff other than the virus, so it could end up being a waste of EA time.

For reasons explained here (https://www.metaculus.com/questions/3711/how-…), I now tentatively believe that Indonesia has an undetected outbreak upward of 2,000 infections. This seems relevant for the question of how much warmth slows down transmissions (the average temperature in Indonesia has been above 30 degrees Celsius).

— edited by ninjaedit

My lower 25% are now at 0.88% (1% previously) because I realized that the age adjustment for the cruise ship study makes a larger difference than I had initially thought. (At the same time 4 more patients from the cruise ship have died since I last looked at this, so the downward update is only small.) And I also realized that South Korea didn't test all the members of the Christian sect, they only conducted interviews with all of them (I'm sure they did a ton of testing on them too, just not everyone). (But even here, South Korea's naive CFR is still ri...

I tried to break down my estimate into subcomponents. **Estimating the total case fatality rate (tCFR)** This number includes asymptomatic cases or cases with mild runouts, and it also factors in changing hospital conditions and so on. Firstly, for **current** conditions (i.e., the conditions where the majority of diagnosed cases are from Hubei, with overcrowded hospitals), my probability distribution currently looks something like this: 70% for 0.5% 20% for 1% 8% for 2% 2% for 3% Conditioning on a world where >100 million people were infected ...

@notany The paper explicitly says they did not factor in hospital overstrain, which would double those numbers in the median estimate, IMO.

@(Uncle Jeff) You and many others seem to be drawing the wrong conclusions from this. I've seen government officials and the person on the most recent Sam Harris podcast on Covid-19 make the same mistake. It's scary how this misconception keeps persisting. This is bad because governments are going to make policy decisions based on this! The problem is this: What the above link says about the cruise ship situation is true. But it isn't news to anyone who has studied fatality estimates carefully. That info was available from long ago and it's already fa...

The numbers are in: 5021 crimes with all categories selected.

I saw this 1h ago. I'm sure the numbers are newer than 40h, but not sure when they were added exactly.

— edited by ninjaedit

@Jgalt And if you had those kind of problems as a country, you probably wouldn't send your doctors to Italy just to do a press conference there saying "You call this a lockdown? I saw someone use public transport?!"

@alexrjl same. I'm stoked!

Uff, two more days until the questions open – almost can't take the wait!

— edited by ninjaedit

Why did this resolve on April 6th? Unless there's a particular reason for April 6th, I think it should resolve April 15th = shortly before the actual announcement of the numbers.

Personally, I have pretty narrow confidence bounds about IFR considerations, so I will tackle this question from this angle: - What's the IFR for the US's population? - %age wise, how many infections progressed to death already? - How many US deaths have been reported so far? - How many Covid-19 deaths so far went unreported? With those four numbers I'll be able to arrive at an estimate for total infections. I can use the serology estimates for NY city for sanity checking. I expect the IFR for the US to be around 1%, and I'm assuming that about ...

@Tamay Maybe most people have had the virus >5 times.

@Jgalt My understanding is that deaths would have to be confirmed to be due to SARS-Cov-2. If I think that most deaths are going to happen in countries with poor health systems, the estimate could be consistent.

@(Jgalt) I don't buy lower than 0.5%. Everyone seems to be repeating the argument that there are many mild cases, but this was already considered by early estimates such as this one (https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-2019-nCoV-severity-10-02-2020.pdf). In the linked report by the Imperial College, their 95% confidence interval does not go below 0.5%. Yes, those numbers might be somewhat dated by now. Still, I get the impression that none of the people who mention numbers lower than 0.5% a...

Made my prediction more pessimistic because the argument that current CFR estimates already factor in hospital crowding to a large degree now seems less compelling to me. Yes Wuhan hospitals were really crowded, but the outbreak was slowed to a halt before more than 5% of the city was infected. In addition, doctors and health experts from other areas could help out in Wuhan. This won't be possible in a global pandemic. I already corrected for the intuition that "things could get worse," but I now think I didn't realize that this consideration could dwarf...

Why predictions above 1.3M? There seem to be only 20 days left and even with a very steep doubling time I don't see how the count would reach 1.3M. Maybe if Iran suddenly reported 500k cases, but that seems unlikely because it would mean switching to a very different way of counting. Oh, NVM. I really need to become better at reading questions before answering. :P 2021

— edited by ninjaedit

@steven0461 crossing the median is for losers (it's literally admitting that you were losing).

@steven0461 Yea but you can just change your prediction to 99% for the rest of the year – that will probably turning losing points into plus. (Although doing that might be a bit ethically questionable! Personally if I got sick I'd change it to 99% but only report once I feel like posting stuff online again, which can be weeks.)

— edited by ninjaedit

People with travel history from Indonesia seem to be showing up as infected in other countries (https://www.thestar.com.my/news/regional/2020…). At least in Singapore, and possibly there was also a case in Australia. If all three examples are correctly reported, the undetected outbreak in Indonesia could well be above 1,000 people by now (my estimate).