James just used the phrase "tail in the distribution," which sparked off something I have been meaning to say. People who use numbers and graphs a moderate amount seem to have an automatic tendency to regard any graph that goes up, then levels off to decline on the other side of the curve at about the same rate. I do it myself now, asI don't encounter that many curves that aren't a Gaussian bell-curve sort of distribution anymore. I used to. There are lots of things in medicine with a long, slow tail, and others that are even more stepwise in deteriorating function. There are graphs which raise sharply, level, then rise again, or dip briefly then rise. Some fall right off a cliff. There are plenty of other curves in the world. I don't think people of low familiarity with numbers and graphs think in terms of any curves at all.
It is the moderately-mathematical, who are numerate but range from below-average to above average dealing with graphs, probabilities, and statistics who are running all sides of the C19 debate. This may be driving the enormous arrogance with which people assert things. They know that they understand math better than most people who can only handle the linear. From this they conclude that they understand more math than those who disagree with them, which does not necessarily follow. Even when it is true - if one actually does know more math than the average journalist or average governor or average attentive political blogger - there is still a good chance you don't know all that much that is useful at present. I can see an enormous amount of popular conversation that overlooks some pretty basic stuff, and I get tempted to make pronouncements of my own of what is likely to happen going forward. Hell, I'd be well above-average at that.
Yet I don't, because I know enough to know that I don't know very much. I confine myself to pointing out over-interpretations, illogical leaps from one place to another, and all the red-flag things I have learned over the years of what is likely to go wrong when people are actually just following sublogical things and then applying their intellects retrospectively to justify their conclusions. According to most research, it's what we do on most of our opinions anyway, hunching an opinion then justifying it post hoc, so I don't feel at all out of line in taking this approach. It's what I can do. I know the danger areas of having a personality that wants the porridge too cold versus wanting it too hot, then going back and "proving" that reality supports our preference for cold.
We've gone over what can go wrong and where those tricks can hide before. Perhaps I will do that again soon, but I don't see that at present. For the moment, I will just say that while our intuitions (and our hearts) leap to a bell curve for the decline of C19, there is no reason to think so. A long tail decline is more likely, and the curve you see for infection when people only take their antibiotic for three days and then stop because they feel better is more likely still.
8 comments:
There's no reason to think that a biological event will perfectly fit any kind of mathematical model. This one didn't on the front side either. It's also really hard to fit America into the model, since we've had this thing in NYC and another thing in Idaho, and fifty or a hundred or a thousand other things. The model is sure to deceive.
That said, there are some practical considerations. The DNC is running a new model that suggests that we could have saved X lives if we'd closed a week earlier. Well, not if it's true that the median age of death is 81, and the majority of deaths are from long-term care facilities and nursing homes, which are by nature effectively closed all the time. The mistake was not taking severe steps to protect those people, rather than that we locked down 17 March instead of 10 March or 3 March. And it may be that people whose median age is 81 are pretty locked down even if they aren't in long term facilities.
My mother went out today for the first time in months and bought flowers. She wore a mask, and the nursery is in the open air. It's a risk; but hopefully not a crazy one, and it will be all right.
Re: the frontside and exponential curves:
https://twitter.com/ARtweaker/status/1264444194178822144?s=20
Grim..."The DNC is running a new model that suggests that we could have saved X lives if we'd closed a week earlier."
Good point about the nursing home factor. Also, if this is based on the Columbia model--then their X lives (36,000) is calculated at a specific recent point in time. That's not a valid way to look at it, because closing earlier can simply push many of the deaths further out rather than eliminating them entirely.
In two years, all of this will seem obvious to us. And in 100 years, it will seem blindingly obvious to those who follow what we should have done to better address the issue. But in the moment, we really don’t know where the data is going to end up. Anyone who makes definitive conclusions at this point is fooling themself.
They are likely on to something, as fewer cases even establishing in the US would be fewer breaches and less contagion even making it near nursing homes. But there is so much missing from any such estimate. Had it been a week earlier, would we have taken exactly the same steps? A little more mask insistence, or marginally less distancing changes the graph. Shutting down arrivals from China and Europe are likely the biggest interventions, and earlier would have been better, later much worse. We know that know, but who knew it then? It was shocking and seemed extreme even when it did happen.
Serious alt-history buffs and wargamers confine themselves to things that could actually have happened. It is a hard discipline, as our memories of chronology, and especially of what we thought about it, are quickly modified to fit what we think they should have been.
As for it being obvious, Granite Dad, I think I can give only partial agreement. Some things will be obvious. Yet some others will seem obvious but be wrong, having entered our folklore in the same way the Great Depression has. The first revisions of the Vietnam War by serious historians are just now being written, and even this far out many are under criticism because the older historians, my generation and the next, still care very much about preserving their narrative because of how it plays out in current culture and politics. No matter what the scientific data tells us in 2070, there will be people absolutely determined to give no credit to Donald Trump about anything.
Here’s some evidence that viral load is more important than age, at least in the developing world:
https://www.washingtonpost.com/world/the_americas/coronavirus-brazil-killing-young-developing-world/2020/05/22/f76d83e8-99e9-11ea-ad79-eef7cd734641_story.html
But the developing world can’t just lock down; not just because of poverty, but because homes often are unsanitary places in poorer parts of poorer countries.
It is not yet proven by any means, but I have been thinking for over a month that viral load is likely important.
For the developing world, people have to put themselves at risk to make any living at all. They don't have the choice, as you note. But the loss of a parent is a devastating driver of poverty in those places.
Not yet proven, yes... but viral load makes a lot of sense to me and I'm structuring my precautions based on that. I also recognize that I'm damned privileged to be able to do so.
Post a Comment