Bottom line - any forecast using "estimated infection and fatality rates" should not be taken seriously.
Originally Posted by lustylad
Well yeah, it has turned out that way hasn't it. Initially people only had data from China, which was suspect. And the Chinese were mostly testing people with symptoms. The first decent indication came from the Diamond Princess cruise ship. Oeb has linked to this. There were 3711 people on board and 700 were infected. Three thousand were tested. So far eight have died. The people on the cruise ship were older than average. Epidemiologists took the fatality rate of a little over 1% and some adjusted this downward for the general population.
As to infection rate, you can't really extrapolate from a cruise ship to the general public. Epidemiologists try to come up with a number they call "R naught", or Ro, being the average number of people infected by a carrier, from testing and tracing. That number is believed to have initially been about 3.5. One person who's infected will in turn infect 3.5 people on average. The number has gone down as people have practiced social distancing, and maybe as some have developed "herd immunity". When Ro goes below 1 and stays below 1 the disease stops spreading and dies out. It should go below 1 when 50% to 75% of the population has had the disease, because of herd immunity. It will go below 1 if everyone shuts themselves away and no one goes out in public, at least as long as they stay shut in.
The infection rate of the population will depend on how Ro changes over time. Predicting changes in Ro must be very difficult, especially when you don't have much testing. Therefore estimating fatalities from the infection rate and the fatality rate are difficult. Epidemiologists came up with everything from 10,000 deaths (low example described in article Oeb linked to) to 2,000,000 deaths (Imperial College estimate, assuming no social distancing, which scared the crap out of people.)
Using historical data like IHME and curve fitting makes a lot more sense, once you've got that data. People can criticize them for dropping their initial estimate from 81,000 deaths to 60,000 now, but either number will be a lot closer to the truth than 2,000,000 deaths.