Alright, gather up. I know all of us were into full-on doom & gloom last week. But I also know that we all kind of settled into social distancing now. Let's talk math, epidemics, and epidemiology. I'll keep it short.
A lot of people have been modeling the epidemic spread as an exponential.
It isn't. It's a sigmoid.
It isn't. It's a sigmoid.
What's a sigmoid function? It's a function that looks like an S that has been slanted to the right. It starts out fast, like an exponential, but then slows down as it approaches an invisible horizontal line.
Look at the number of cases in China, the red line. It's a clear sigmoid. 
Epidemics are sigmoids because the pool of susceptible people dwindles down over time. As the disease infects more people, it becomes increasingly difficult for a sick person to encounter and infect a previously uninfected person.
The invisible horizontal line (asymptote) that the sigmoid approaches depends on just two big things:
1. Whether R0 is below 1.
2. The percent of vulnerable population.
But that graph also depends crucially on another:
3. The amount of testing.
1. Whether R0 is below 1.
2. The percent of vulnerable population.
But that graph also depends crucially on another:
3. The amount of testing.
As a country increases testing, the number of cases will go up. Sometimes very fast, as they deploy more tests. One cannot interpret the daily numbers unless one compensates for changes in the number of tests administered and testing protocols.
How many of the techbros who opine on this virus actually follow any of this?
What's the maximal percentage of the population that's susceptible to the severe forms of sars-cov2? No one knows. There are quite a few genes that confer protection against related pathogens. Thalassemia trait, blood types, and other genes may well prohibit severe forms of it.
The cruise ship provided an interesting data point: of the 3711 people on board, only 715 got covid-19 at all. Far fewer had the severe form of the illness. The asymptote was fairly low.
So that line isn't at 20% (the percentage of severe cases) of 100% of humanity. It's possibly far lower. And with social distancing, contact tracing, extensive testing and quarantine, we can lower it still.
Do not pay attention to any epidemic models which model R0 as a single number. Infection isn't uniform. We have superspreaders: most people only infect a small number, but some infect large groups, possibly because of certain social interactions. These folks pull the average up.
For instance, superspreading occurs at religious congregations, funerals, crowded clubs and the like. Control those, and R0 is reduced drastically.
There's a reason we have professionals who study public health. There's a reason why we train epidemiologists and don't just hire techbros.
The numbers involved will seem very large, because they are large. But hardly anyone puts them in context. We had one guy predicting that 1.5m will die globally. Ok, but 60m people die globally every year. That number is bad but not that bad, esp given there will be overlap.
A news channel that recently divided 500m by 317m and somehow managed to get 1.53m as a result is now spreading fear about the virus. Innumeracy is everywhere.
It's definitely Not Just Another Flu, but it's also not The End Of The World. It's significant, requires a drastic change in lifestyle, but it's controllable. We know this because we've. observed. it. get. controlled. in. multiple. countries. Just look up at the graph above.
I'm just another techbro, albeit one who works on epidemic algorithms, like the Avalanche consensus protocol. These protocols are governed by sigmoids.
My main message is simply to say: do not listen to me or others like me. We lack the credentials/knowledge.
Do not follow the causalty numbers, except as a rough guide. We lack the context.
Do not pay much heed to the online simulators. They lack precision.
Do not follow the causalty numbers, except as a rough guide. We lack the context.
Do not pay much heed to the online simulators. They lack precision.
Follow the actual experts. There's a lot of noise online, backed by a lot of flawed math and flawed reasoning. The actual experts have proper models, where they simulate actual people engaging in real life activities. They can make proper predictions.
There are only 300m Americans. The government has extensive data on all of us. It's not computationally difficult to track the disease on an individual basis.
Don't give into the doom & gloomers online. Demand a science based policy from politicians. Otherwise follow the sensible measures that every sane country is converging towards. We'll get over this.
