Arguing with what I thought was a complete idiot on Twitter thinking he's just some rando and it turns out he's a game programmer with 100k followers, went to UC Berkeley. I mean, it's a puzzle game, surely he can logic, but my mind is kind of blown by how he can program his way out of a paper bag given the lack of rigor in his thinking.
Basically, his argument was that the fact that Li Wenliang died, and was only 33 years old, meant that the overall fatality rate for everyone infected being 1% - which was estimated by a team of expert epidemiologists and based on real actual data- was too low.
Anyway, it's an interesting problem to figure out how his cognition went astray. I think it went something like this:
* News says mostly killing older people.
* News says famous doctor died of it. Only 33. This contradicts mostly older people thing.
* The fact that I know one 33 year old died means that 33 year olds dying is an expected and probable event.
* A probable event (33 year old dying) being less probable than an old person dying must mean and old person dying must be much more probable than I previously thought.
* I previously thought it was (5%-10%)?
* Ergo 1% is much too low an estimate for infection mortality rate.
In case you aren't aware why this is a completely bad and wrong train of logic, well there are a lot of reasons.
Li Wenliang being in the news and being 33 doesn't actually make this 33 year-olds dying more likely. This is a cognitive bias we all have; if we see something in the news we mentally raise its probability. This is often wrong. It's called the availability heuristic. https://www.verywellmind.com/availability-heuristic-2794824
But let's assume, for the sake of argument, that Li Wenlian dying means we now know the infection fatality rate is 10% for 16-59! And we know older people is higher, but we don't know what it is, so n >=10%. Let's assume 0-16 fatality rate is 0%. And we know population demographics looks like this: 10% 0-15, 60% 16-59, 30% 60+. So we can say total infection fatality rate is between 9% and and 36%. A ha! His logic works!
Except the thing is don't actually *have* numbers on what percentage of people are killed by age, because age at death and ... death... is collected at the same time. Estimating total case fatality rate by using age specific case fatality rate is putting the cart before the horse! These estimates for fatality rate for specific age brackets are ALWAYS worse than the total fatality rate for all ages because you lose statistical power. When we have age specific fatality rates, we would calculate this using the SAME data we ALREADY used to estimate the infection mortality rate of 1%. But the data aren't good enough for this yet.
So basically, his train of thought requires us to live in a world where we know age specific mortality rate *before* we know the total mortality rate, and additionally, requires us to live in a world where "improbable" things, like a 33 year old famous doctor dying of the disease he's known for discovering, never happen.
The fact is that being famous doesn't stop you from dying in improbable ways. And things with a probability of 1% happen to us all the time. IUDs are >99% effective and I have a friend who got pregnant with one in! 1% may seem small to us, but 1% of a million is 10,000. Swine flu had a really low infection fatality rate of .01% and it left 200,000 people dead, because it just infected so many people. Improbable things, applied to a large enough population, become probable, and then they become certainty. You just have to hope *you* aren't the statistic, because someone out there always is.