I do not like the way that the experts are dealing with this crisis. I am at the point where I don’t care about being charitable toward them. Nassim Taleb, who is never charitable to those who disagree with him, has a shorthand that I will modify for this purpose. Expert Yet Idiot, or EYI.
- What should we rely on to make decisions?
- ASK: rigorous studies and experiments
- EYI: noisy data and models
- What should we use to reduce the spread rate of the virus?
- ASK: masks and scarves
- EYI: lockdowns
- What should government do to relieve individuals and small businesses?
- ASK: provide backing for credit lines from banks
- EYI: Massive intervention by the Fed and deficit spending
Regarding (1), we need to have an idea of the prevalence of the virus in the population. It is impossible to do this by following the data on reported cases. The amount of testing varies day by day. Different tests are used, with different rates of false positives and false negatives. The lag in reporting results can be anywhere from less than an hour to more than a week. Tests are done on non-representative samples of the population. A rigorous study would use a proper sample design and strive for uniformity in the time and method in which tests are done.
Also, there is no definitive view of how this spreads from one person to another. Experiments would be getter than guesswork.
Models do more harm than good. They rely on data and assumptions that propagate through the model, giving precise-seeming results that are in fact unreliable by orders of magnitude. UPDATE: Peter Attia has a great 8 minute YouTube on this.
For now, the only forecasting method I trust is to extrapolate the growth rate in the number of deaths for a few more weeks. Until today, deaths were doubling very steadily every three days. Today, the ratio of deaths to the total three days ago dropped to 1.86. That is the best number to track until we get something more rigorous to use.
Regarding (2), it seems intuitively plausible that lockdowns will reduce the spread rate. But it also seems intuitively plausible that having everyone use masks and scarves would lower the spread rate. Neither approach will drive the spread rate to zero, but perhaps either approach would stop the phenomenon of the superspreader–one person whose case is such that it infects many people.
The difference between masks/scarves and lockdown is that maybe I could get a haircut. That is if everyone in the barber shop had masks and scarves, and if the barber took unusual measures to clean the chair and his tools.
Regarding (3), if you have not been following this blog, scroll back through previous posts.
Not sure what testing gets you even if it’s done to rigorous random sampling parameters. Do you make a different decision if 1 percent have a particular strain of the virus or if 80 percent have it? The relevant information for decision-making seems to be hospitalization. If, because it primarily hospitalizes the old and already sick, someone wants to say quarantine these specific groups, and let everyone else take there chances with scarves and masks, you can make that argument, but population prevalence will be irrelevant, because the bottom line will still be hospitalization case load. Knowing the precise prevalence is not going to make it any easier to estimate the hospitalization case load consequences without additional information about the age and medical conditions of the sample, and the sample will only be good for a few days and have to be repeated multiple times to get trend information. Pie in the sky. Like the idea of isolating positives and tracking and tracing contacts. In what world was that ever going to happen.
Bolsonaro is arguing against lockdown and advocating quarantine of the elderly and is getting savaged for it. Not that it is making any difference in Brazil, an authentic federal republic, where the governors have authority, matter, and are held accountable through genuinely democratic elections. Sweden is doing the same thing but Bolsonaro gets targeted by the racists in the media and academia.
I’m with you on (1) and (3). Experiment. Test. Loan. My disagreement on (2) is that everything I learn vies against the efficacy of masks. For example, I found this quite credible: https://vimeo.com/399733860) I am all for testing more here, as with asymptomatic transmission, but I am increasingly convinced that this is at best a marginal possibility, probably worth less than the cost of the mask themselves. Do you have some evidence to the contrary?
Arnold;
While many things you say make perfect sense, the one about doing careful science – in this case – is absolutely wrong. I hear it regularly as a critique of the current leadership; but it is, in effect, a science fetish. The pandemic is fast enough moving that countermeasures must be deployed by analogy with other outbreaks in the absence of good data, let alone experimentation. Unfortunately, the initial analogies were to SARS and MERS – which didn’t turn out this way – and the initial data was obfuscated long enough that we were near/at pandemic before half-decent observations (basically, European; given how idiosyncratic the cruise ship and South Korean church were) were available.
These past two-three months are 100% the wrong time to do carefully constructed trials or collect random data using diverse, time consuming, scarce diagnostic tests designed for confirmatory testing. If we had been designing screening tests for contact tracing back in December, back when samples weren’t being shared…
Oh, just so we have some context on the models being used:
https://www.nature.com/articles/d41586-020-01003-6
And models do not do ‘more harm than good’ because everyone approaches the situation with a model. If you don’t give them an exponential model, they start with a linear trend line (oops).
You might be amazed at how much a bad model will give you bad information, or how much a good model with bad input will do the same. If models had agency we would say they lie, but at the very least “garbage in, garbage out” always holds true. The problem is that the garbage coming out has been laundered and looks lovely and scientific, but it is still garbage.
“ Not sure what testing gets you even if it’s done to rigorous random sampling parameters.“
80% Infected: Once these people recover we will have reached the threshold for herd immunity where 70%+ of the population has been infected and recovered and has antibodies which provide somewhere between short term partial immunity and long term full immunity. It will mean CV19 is much LESS serious than the seasonal flu.
1% Infected: We are so far from reaching herd immunity this is the tip of the tip of the iceberg with millions of deaths worldwide in the next year. It will mean this could be 10x worse than seasonal flu.
“Do you make a different decision if 1 percent have a particular strain of the virus or if 80 percent have it?”
80%: We are at peak hospital demand and we can all go back to living.
1%: If we don’t maintain social distancing for 12 to 18 months until a vaccine is developed hospitals will be completely overwhelmed and they will be filling parking structures with dead bodies.
Kingtu- Your argument for 80% herd immunity would work only in community conscious, educated and where generations live separately, like Sweden. It does not work in NYC. Much less in Rio de Janeiro with 40% of the population in slums with multiple generations cramped in one room and barely civilized. If you don’t mandatorily isolate the infection rate sky rockets and you get an apocaliptical health care situation which no government would survive. In any case, as we will see in Italy and Spain, the death and prolonged disability rate in the 50 to 70 years old bracket will be very damaging in terms of lost expertise in management and science. If we live through, we will see…
In short, the epidemic is heavily weighted by sociological factors.