In theory. On the other hand, there are no deaths from famine from the Great Depression and that generation ended up displaying a positive spike in longevity.
sir2
The body seems to slow down aging during times of hardship, to give the species a chance to procreate at a later date.
Let’s say that I tell you that if 100% of the population smokes one cigarette a day then the whole population will lose 15 years of life and if no one ever smokes a single day in their life then no years of life will be lost.
What have I predicted? Did I predict 15 years of life will be lost or did I predict 0? Are my numbers completely wrong? That’s the wrong way of thinking about a prediction.
The word “predict” in scientific modeling means a different thing from what you think it means.
A scientific prediction gives scenarios that allow the decision makers to get a flavor of the range of possibilities and what sorts of things will happen on the basis of those. It’s not an attempt to predict the future any more than I am predicting what you will order at my restaurant when I hand you the menu. I can tell you what you’re not going to order, but you have full choice from anything on the menu itself and I can’t force you to pick one, I can just tell you which are smarter and which are dumber.
But now let’s say that the decision makers make a decision and go back to the scientific modeler a month later. Well, now one month’s worth of time has been settled and can’t be changed. Any models of what might happen from here, depending on the choices, are going to have to start from the current position.
If 30% of people have been smoking for the last 20% of the time period that we care about then, no matter what, I can’t get the years of life lost to smoking down to 0. It’s too late. We’re already 20% through and we didn’t maintain the zero smoker result that we were hoping to accomplish. Our range of possibilities will now be smaller - between 3 years of life lost and 15 - and that’s not more accurate a prediction than last time because I was stupid, ignorant, and incapable before, it’s because options have now been taken off the menu and are no longer available.
Now in the case of FiveThirtyEight and their Covid-19 predictions, probably the scientists have improved their models between then and now, to be more accurate. But they’re still not “predicting the future” in the way that you think. They’re giving a range of options that decision makers can choose from and saying where we should expect to land, on the basis of those. If you see the decision makers making good choices, then we won’t land “somewhere in the middle” of the predictions, we’ll land at the bottom end of them, because that’s where we chose to drive. If we land in the middle or at the top, it’s because the decision makers decided to be morons and drive towards more death.
As said, it’s a menu, not a prediction.
We end up inside the bars based on where we’re mentally capable of arriving. It’s not prognostication, it’s our range of options.
I see a report of China’s industrial shutdown’s health benefits.
I doubt such benefit can be obtained in the US even with newly-relaxed pollution limits. Fewer factories and vehicles are spewing but they can spew more. That likely balances.
The fact is, you haven’t a fucking clue how these experts came up with their estimates. And I know perfectly well in what way they’re “predicting the future” and have no idea why you would think I didn’t.
We briefly discussed this group of experts in the predict the fatalities thread. They are no doubt experts in epidemiology but I really doubt they’re all intensely involved with building up to date national models. For reference:
Eta: but I bet that first guy totally is.
There are about 2000-2500 intimate partner homicides per year, and another 1500-2000 child deaths from abuse or neglect in the US. Even assuming these numbers are under reported, it is important to keep them in perspective compared to COVID-19 deaths. As of this posting, that number is about 14,300. Domestic abuse deaths could go up by 10 fold, and still not reach even the lowest estimates of total deaths from COVID-19.
So, trolley problem: On the track ahead of you are three people that will surely be killed if the trolley continues. The trolley can be diverted to another track, but that track goes around a corner, and you can’t see what is on it. Do you divert the trolley? What if you know that sometimes there are people on the tracks around the corner?
It’s a interesting idea, but maybe in the long term. I read that our young men fighting in Iraq and Afghanistan overseas actually have a lower death rate than if they stayed here in the States.
Less gang wars, less auto accidents, and first rate super fast medical care.
If Covid gets a significant number to quit smoking, that could, indeed- in the very long run- reduce early deaths.
I agree that there will fewer auto accident deaths, but I doubt that will equal out the deaths from the virus.
Last one on the list:
Srinivasan (Srini) Venkatramanan is a research scientist at the Network Systems Science and Advanced Computing division. Prior to joining University of Virginia in 2018, Venkatramanan did his postdoctoral research at the Network Dynamics and Simulation Science Laboratory, Biocomplexity Institute of Virginia Tech (2015-2017), where he also worked as a Computational Health Data Scientist (2017-2018). He received his Ph.D. in Electrical Engineering from the Indian Institute of Science in 2014 for his dissertation titled “Influence Dynamics on Social Networks”. Venkatramanan’s research areas include stochastic modeling, diffusion dynamics, optimal control and network science. At the Biocomplexity Institute, he is responsible for developing, analyzing and optimizing computational models for complex systems arising in the domains of epidemiology and food security.
Being cranky for the sake of being cranky is a worse mechanism for modeling and judging modeling than computer simulation.
Very good observations. To quote a comic I heard on NPR said, in reference to doing lots of cooking and eating while sheltering in place, “hey, I gotta eat this - those pounds aren’t going to gain themselves!”
And, even if the “heavy fast-food users”* aren’t going to McDonald’s for every meal now, they’re instead eating all of the processed food they stocked up on at the grocery store (lots of Spam, SpaghettiOs, and frozen pizzas).
Moderator Note
Dial it back. No warning issued, but keep it civil in this forum.
Colibri
Quarantine Zone Moderator
I didn’t question the credentials of anyone on that list, I just doubt they are all informed daily on new data to inform their estimates with.
This started because you said 200k was a low end estimate. Point me to some real model being put out by a reputable organization. A bunch of experts who get to anonymously make guesses is not that. This isn’t being cranky for kicks. Your own cite said that 70k was a reasonable guess too. Here’s a projection from IHME at the University of Washington:
It has it at about 60k deaths. I am not endorsing here, just reiterating that 200k deaths is not a low end estimate.
I have a very good idea.
If someone said, “Well, you know, purple is the complementary color of blue.” And it seemed to be necessary to go over the additive and subtractive color wheels with them, then I would be skeptical of their self-proclaimed mastery of art science.
Mixing up worldwide statistics with US is fair. Ignoring that, though, you’ve also proclaimed that a bunch of computer epidemic modelers didn’t do any simulations to make their predictions and said that their predictions were “wrong” when talking about people presenting the scope of options. Unless the new scope lies outside of the range of the previous ones - which is unlikely - the previous predictions were still accurate. We’ve just narrowed in on the correct range.
So you’re standing firm that 200k is a reasonable consensus low end estimate?
That’s probably reasonable. But, likewise, when 538 updates the Presidential tracker, they usually get in giant-and-authoritative surveys at extended delays and mini-surveys which are to be held to be less reliable at smaller intervals. The best output from these two styles of data is an intelligent mixture of all, based on their relative merits and timeliness.
I would need to review the age of the models being included in 538’s panel, but I would expect that they’ll toss any which have fallen too far behind the times. But I wouldn’t expect them to throw away everything that wasn’t made fresh for today. That would probably be too noisy to be useful.
Unless you see some better source of data to turn to. Citing the experts is the best that can really be done, whether you’re happy with their output or not. They’re still the experts, even if the science is in its infancy.
20 experts gave their best go at offering some models and that’s where they landed. I could create my own simulator but I would only have the possible benefit over them that I would be seeking to perform an actual, colloquial “prediction” on the basis of my own personal sense of cynicism and optimism about what choices humanity will make. The end result would be a judgement of where I lie on that metric and how close I am to understanding the world. If it’s right then I would have a “better” prediction than theirs. But it wouldn’t be so useful for a decision maker as what the professionals do.
Which expert gave the best answers, do you think?
Imagine that I give you and nineteen other people the task of modeling a sphere. However, you can only use triangles to do it and you’re only allowed to use 12 triangles.
Which of the 20 of you will have the most accurate model? Likely, none. If I wanted to reconstruct the shape that all of you were going for, my best bet would be blend all twenty results together and hope that different low-resolution views on the same object, when melded together, reveal a higher-resolution result.
OK, now you imagine 20 meteorologists being asked to predict American weather a month from now. Further imagine there’s a brand new jet stream they’ve never seen before.
There’s a bunch of reasonable models from reputable people out there. 200k is simply not a low end estimate. If the 538 article pegged specific estimates to specific experts, I could be suitably more impressed. If the guy who’s resume you quoted was speaking, I’d certainly listen. But was he one of the low guesses or high? We don’t know.
I suspect that you can track down each of their papers and evaluate. But if the first guy was great and the last guy was great, the people who selected them are experts in statistical analysis, and the results are all pretty in-tune with one another, I see no big reason to get finicky about the subject.