I reiterate that people are not making enough predictions about the situation. I made some predictions 2 weeks ago. I had a couple hits and a couple misses. See post two weeks ago here:
https://steemit.com/…/@pasha-…/basic-coronavirus-predictions
Hits:
South Korea:
predictions 10000, actual ~9000
Numbers looked like they were slowing down.
Italy:
Prediction 66000, actual ~60000
Again a strong prediction especially given that Italy had 7000 cases 2 weeks ago
Somewhat misses:
Iran:
Prediction 11000, actual 20000
Iran's numbers are fake and it's unclear which curve they want to be seen as following. 2 weeks ago, the raw numbers were slowing down, but it seems somewhat hard to predict based on that alone.
USA:
Prediction 10000, actual 33546. This is somewhat of an interesting lesson. I originally applied the Italy growth model and got 27000 for the US, then decided that was too much and tried the China growth model, giving 10000. Note the US had 500 official cases 2 weeks ago.
Also note that infections decease experts struggled quite a bit with US predictions as well
They expected the US to grow from 3.5k cases to 19k cases from March 15 to 29th, while we already have 33k+ on March 22nd. In other words, the average group consensus (usually better than a lot of individual predictors) was about 9 days off in US, while I was 4 days off in US. (I was 9 days off in Iran, but that was my worst prediction and my average is quite good).
So the new predictions for the next 2 weeks (April 5th)
South Korea:
South Korea is stable and will continue to be so, while still dealing with a trickle of imported cases:
Prediction: 10000
In addition to the prediction I am going to add an “outside view” bound. Last time even on the wrong cases I was off by at most a factor of 3.3, so I expect at most this error now.
Bound: (9000 - 30000)
Italy:
Italy has just hit a logistic middle today (between 53.5 and 59.1 thousand), with an average of 56.3. So I will expect it to stabilize at around 112 thousand.
Prediction: 110000
Bound: (70 000 - 350 000)
Iran:
Iran’s numbers are fake, as usual but it seems that the numbers are very linear (or mostly linear with a slight quadratic trend). It’s kind of interesting that “quadratic trend” is something we observe I countries when we suspect they are lying and “logistic trend” is something we observe if we think they are telling the truth. If it try to fit a quadratic plot for the numbers from Feb26th to March 21st, I get a prediction of 46200. If I try to fit a linear plot, for the numbers of March 6th to March 21st, I get a prediction of 36800. If I try to take their fluctuating point as a sign of half-way logistic regression, I get a maximum bound of 43200. Taking a rough average of those, I get around 42000
Prediction: 42000
Bound: (30 000 – 130 000)
USA:
USA’s numbers are all over the place, so using logistic estimation seems flawed at this point (it will become more correct in the future). The main lesson from last prediction round is that the best model for the US is looking at per-capita model for Italy and seeing how they match up.
It seems that Wikipedia is having a weird issue with the numbers as the numbers for the US keep going up and down for today. I am going to use https://www.worldometers.info/coronavirus/country/us/ for the latest number and Wikipedia for the other ones.
We are at 33546 cases. US population is 327.17 million, Italy population is 60.80 million. US is roughly 5.4 times larger, so the equivalent per-capita Italy cases was 6212, which they reached between March 7th or 8th. Match 7th to 21st was a 53578 / 5883 = 9.1 increase, so I am expecting the US to have a similar increase, which means: ~306K cases. Originally I was worried about testing capacity, but this will likely to not be an issue.
Prediction: 306 000
Bound: (90 000 – 900 000).
Quick predictions on other countries. Looks to me that Spain, Germany and France have all hit their halfway point in the last couple days, so predictions for them:
Spain:
Stability at 58 thousand, 55 000 in 2 weeks (bound 35 000 - 170 000)
Germany:
Stability at 42 000, 40 000 in 2 weeks (bound 30 000 - 130 000)
France:
Stability at 34 000, 32 000 in 2 weeks. (bound 20 000 - 100 000)
Summary:
South Korea: 10000 (9000 - 30 000)
Italy: 110 000 (70 000 -350 000)
Iran: 42 000 (30 000 - 130 000)
USA: 306 000 (90 000 – 900 000)
Spain: 55 000 (35 000 - 170 000)
Germany: 40 000 (30 000 - 130 000)
France: 32 000 (20 000 - 100 000)
Is there good news?
Yes! I think the European countries will begin to stabilize in 2 weeks. I think the US cases will grow but it’s possible that the US will hit the halfway point in 2 weeks and begin to stabilize 2-4 weeks later at around 600,000 cases. So, I think we have about a 4-6 weeks more of hard lockdowns at which point we would be able to restart parts of the economy while EVERYONE WEARS MASKS. If everyone does not wear masks, then the whole after – lockdown thing fails and we either have another massive wave right away or another prolonged shutdown.
This model is an “outside view model,” rather than a ground up model. It’s compatible with other outside view models of behavior. In simpler terms, this means the model already assumes that WE ARE GOING TO TIGHTEN RESTRICTIONS AND SOCIAL DISTANCING and that we are going to find drugs that work somewhat.
Wish everyone the best.