Why No COVID-19 Models Have Been Accurate, And How To Fix That

[TBoneTX 45,571]

Comments, especially from a statistical standpoint?

Quote

Why No COVID-19 Models Have Been Accurate, And How To Fix That

by Jon McCloskey - April 27, 2020

 

[...] The problem with data-driven models, especially when data is lacking, can be easily explained.  First of all, in terms of helping decision makers make quality decisions, statistical hypothesis testing and data analysis is just one tool in a large tool box. [...]

 

Continues here: https://thefederalist.com/2020/04/27/why-no-covid-19-models-have-been-accurate-and-how-to-fix-that/

 

[Dashinka 24,628]

This really is key.  I have used models for a lot of work, and the more complex a system is, models become less and less helpful.  You can think of what is done around climate change.

 

Considering the limitations of this tool under controlled laboratory conditions, imagine what happens within more complex systems that encompass large areas, contain millions of people, and vary with time (such as seasonal or annual changes). In fact, for predicting outcomes within complex and adaptive and dynamic systems, where controlled experiments are not possible, data is lacking, and large amounts of uncertainty exist, the reductionists’ tool is not useful.

[TBoneTX 45,571]

Oh, but climate change is "settled science."  

 

Most disturbing to me regarding the Virus protocol are these 4 (or 3.5) issues:

  -- continuing to operate under the initial "disastrous" models in the face of real-world evidence

  -- continuing to operate as though all the U.S. is like New York City

  -- moving the goalposts from "prevent deaths" to "keep absolutely everyone from getting sick"

  -- quarantining the healthy (people & the economy) instead of the sick and the highest-risk groups

[Dashinka 24,628]

The funny thing is the impacts to the economy relative to experiencing a depression will be what historians will talk about a hundred years from now in my opinion.  Chairman Xi’s bat cold will be be forgotten.

[90DayFinancier 910]

Any criticism of the economic impact of Corona Virus control and mitigation, that ignores the cost of doing nothing, both in human and in economic terms is cognitive dissonance to the highest degree. 

 

Anyone can take shots from the cheap seats, is people with medical and economic solutions that we need to listen too. 

[Burnt Reynolds 5,058]

This is a great read.

 

The topic in this article not only goes for models but also goes for business decisions. In management we're bombarded with the usual analytics and inputs, whether from executives (especially new/younger ones) above or just from disconnected stakeholders. I'm under the impression that there's this pressure from (younger) executives because of academia trying to dictate this pointless and moronic model thinking when in most cases it's irrelevant never mind harmful. Whether it's climate models, virus spread models, or algorithmic models (what Twitter likes to blame when employing new methods of censorship), or internal analytics and modeling, they always need humans making end to end decisions.  

 

I've seen a couple areas where modeling is useful and quite accurate, but it has very high consistency and homogeneity (not to mention sound methodology), whereas models like the virus spread, deaths, climate and so on have so much complexity it's just being proven pointless. It's again like lie detectors where the real world use of them is functionally to misdirect.. rather than being mediums of fact or truth they're policy instruments by people wanting a certain outcome, or just someone who wants to look good screwing around with modeling, like the magician performing magic tricks getting ooh's and ahh's. It's great if you want to toy around with R, or Markstrat, and so on, but has no basis being seriously considered in reality. The former, however, it what's been guiding political policy and it's become highly destructive to western political discourse, not to mention destroying science.

[Dashinka 24,628]

33 minutes ago, 90DayFinancier said:

Any criticism of the economic impact of Corona Virus control and mitigation, that ignores the cost of doing nothing, both in human and in economic terms is cognitive dissonance to the highest degree. 

 

Anyone can take shots from the cheap seats, is people with medical and economic solutions that we need to listen too. 

No one suggested doing nothing, there are other strategies that would have accomplished the same goal as flattening the curve without such drastic impacts to the economy.  The Covid will work it’s way through the system in a relatively brief period of time.  There may or may not be a vaccine at some point, hopefully we will get better therapeutics, but the economic impact will be felt for years at this point.  We haven’t even begun to see the massive money printing that will be going on to appease the states, and there will be a tremendous loss of life from many more factors due to the failed economy.

Edited April 27, 2020 by Dashinka

[90DayFinancier 910]

1 hour ago, Dashinka said:

No one suggested doing nothing, there are other strategies that would have accomplished the same goal as flattening the curve without such drastic impacts to the economy.  The Covid will work it’s way through the system in a relatively brief period of time.  There may or may not be a vaccine at some point, hopefully we will get better therapeutics, but the economic impact will be felt for years at this point.  We haven’t even begun to see the massive money printing that will be going on to appease the states, and there will be a tremendous loss of life from many more factors due to the failed economy.

You paraphrased my statement incorrectly

 

"Any criticism of the economic impact of Corona Virus control and mitigation, that ignores the cost of doing nothing, both in human and in economic terms is cognitive dissonance to the highest degree. "

 

 

[TBoneTX 45,571]

6 hours ago, Burnt Reynolds said:

This is a great read.

Oh, but it happened to be published on The Federalist, so therefore it must be dismissed out of hand.

[OriZ 3,157]

Personally, I feel pretty good about the model/s I've posted starting a couple months ago vs what actually happened.

 

Regarding other models, I think some were more accurate, some less - the biggest sticking point to me with the ones that I thought were way too pessimistic(pretty much anything Neil Ferguson has anything to do with), is that they all extrapolated totally linearly, and failed to account for several factors such as weather. Math is great, but basically it comes down to I think most of it was just lazy. 

[Nature Boy 2.0 8,376]

Modeling is not rocket science if you have multiple similar past occurrences to compare.

 

As a hobby I keep up with all kind of data on sales. I am sure managers can mine it any time they want. I dont have that access so I just keep up with it. 5 years worth.

 

My coworkers always ask me what the month will look like. Sometimes I can tell them, save every dime its going to be a ruff month.

 

I can tell you December will be huge but October and November will suck.

 

The problem modeling this is we have no history 

[Steeleballz 7,550]

50 minutes ago, Nature Boy 2.0 said:

Modeling is not rocket science if you have multiple similar past occurrences to compare.

 

As a hobby I keep up with all kind of data on sales. I am sure managers can mine it any time they want. I dont have that access so I just keep up with it. 5 years worth.

 

My coworkers always ask me what the month will look like. Sometimes I can tell them, save every dime its going to be a ruff month.

 

I can tell you December will be huge but October and November will suck.

 

The problem modeling this is we have no history 

 

   

  We can probably get a good idea with what happens in southern hemisphere countries over the next few months.  

[Dashinka 24,628]

10 hours ago, OriZ said:

Personally, I feel pretty good about the model/s I've posted starting a couple months ago vs what actually happened.

 

Regarding other models, I think some were more accurate, some less - the biggest sticking point to me with the ones that I thought were way too pessimistic(pretty much anything Neil Ferguson has anything to do with), is that they all extrapolated totally linearly, and failed to account for several factors such as weather. Math is great, but basically it comes down to I think most of it was just lazy. 

I agree with you, but the sad thing is a lot of impactful decisions were made based on what Neil Ferguson said.  Even the UW model that I find a lot better needs to be updated more often than what they do.

[OriZ 3,157]

4 minutes ago, Dashinka said:

I agree with you, but the sad thing is a lot of impactful decisions were made based on what Neil Ferguson said.  Even the UW model that I find a lot better needs to be updated more often than what they do.

I honestly don't know why anyone would listen to him after his calls about Mad Cow and other stuff were so off. I still supported the measures taken to flatten the curve despite of that, but not because of it.

 

 

[Dashinka 24,628]

1 minute ago, OriZ said:

I honestly don't know why anyone would listen to him after his calls about Mad Cow and other stuff were so off. I still supported the measures taken to flatten the curve despite of that, but not because of it.

 

 

I was fine with the measures to flatten the curve and take the stress off the medical system.  My main beef is with the changing narrative of the purpose of all these measures.