- We already have 4 other Corona Viruses endemic in the human population
- More than 80 percent of confirmed coronavirus cases are not severe (probably because T-Cell induced immunity is working very efficiently)
- Reported reinfections are minimal
I think these priors alone reduce the probability for ADE drastically. Don't you?
I was under the impression that virus formation (copying, mutations and all that) was indeed a statistical process. But maybe I am wrong?
For example, one can assume a gaussian distribution and estimate its mean and width using the mean and standard deviation of the data. Bayesian inference aims to directly extract the underlying probability distribution without assuming a fixed functional form.
In any case, one should be extremely skeptical when someone tries to apply such methods to sparse or cherry-picked data. Including SARS-CoV-2, there are only seven coronaviruses that are known to infect humans. This is an extremely small number to do statistics with.
Of course one can make educated guesses based on experiences with previous similar viruses (such as SARS-CoV). But using the language of Bayesian inference for this is just cargo cult science in my opinion.
Pretending that human reasoning is somehow equivalent to Bayesian inference (and abusing domain-specific terminology like "prior") is exactly the kind of cargo culting that Feynman wrote about. If there is no actual mathematical model and software implementation, you are not doing Bayesian inference (and the airplanes will just not land).
Should we ignore everything we know from the past?
As in, looking behind you