This Could Be a Clue as to why Minnesota’s COVID-19 model was wrong by at least one, and perhaps two, orders of magnitude

Powerline Blog ^ | 05/16/2020 | John Hinderaker

[SeekAndFind]

This Could Be a Clue…

…as to why Minnesota’s COVID-19 model was wrong by at least one, and perhaps two, orders of magnitude:

Before Friday, March 20, Marina Kirkeide, who graduated from the University of Minnesota College of Science and Engineering in 2019, was a School of Public Health part-time research assistant working on HPV transmission for Kulasingam. On a gap year before starting Medical School at the University in fall 2020, Kirkeide also had a second job as a lab tech at St. Paul’s Regions Hospital. That Friday, Kulasingam called her and two other research assistants and asked if anyone was available to “work through the day and night” to get a COVID-19 model to Governor Walz the following Monday. They all jumped at the chance.

The model was literally created over a weekend, in part, at least, by kids.

“I don’t think a lot of researchers get to work on something over the weekend and have public figures talk about it and make decisions based on it three days later,” says Kirkeide, who had to leave her hospital job to focus solely on modeling. She feels the responsibility of such a big project, too. “[In this situation] you don’t have the time to validate as much as you normally would. You want to get it right the first time.”

They didn’t. Minnesota is now using the third iteration of the U of M model.

As a modeler, says Kirkeide, you have complete control over what your results look like. The most important thing is to have absolute integrity.

“Yes, numbers may look grim, but they are what we’re getting,” she says. “You can’t argue with what you see.”

You can, however, check your model against reality. These are the “grim numbers” the modelers were projecting:

One projection showed that cases would peak around April 26 in Minnesota if there were no mitigating steps to slow the virus. The death toll in this scenario could reach 74,000. The other scenario showed a time frame with significant and staged mitigations in place that pushed the peak to about June 29 and projected deaths in the 50,000–55,000 range.

In the linked article, the University of Minnesota’s School of Public Health boasts that “[w]hen [Governor Tim] Walz issued the stay-at-home order for the state two days later, which he recently extended to May 4, he took these projections heavily into account.” Six weeks later, is is obvious to all that the U of M model was wildly off the mark. As noted above, it has been replaced by later versions that still can’t “predict” the present, let alone the future. But no matter: the policy lives on, long after the basis for the policy is gone.

[alstewartfan]

Somehow, Glenn Beck’s “The true purpose of coronavirus wargames” is still on YouTube. Shocking and illuminating.

[nwrep]

[PeterPrinciple]

The model was literally created over a weekend, in part, at least, by kids.

---

I put together a lot of business plans that would make me a millionaire in a year............................none of them worked.

[SoCal Pubbie]

The infamous Covidiots on this forum have defended Neil Ferguson’s wildly inaccurate model because they claim the actual death toll fits into the lower strata of his predictions with mitigation, or that mitigation was more successful than anticipated. The old heads I win, tails you lose trick. I guess that’s how they’ll explain this away too.

[PeterPrinciple]

As a modeler, says Kirkeide, you have complete control over what your results look like. The most important thing is to have absolute integrity.

---

What is absolute integrity:

Acting in accordance with your beliefs. Nothing to do with truth or science.

https://www.quora.com/What-is-absolute-integrity?share=1

[econjack]

I’m a member of the U of FL software team that is building a low-cost (i.e., under $200) ventilator for Third World countries. It took six of us working around the clock for over a month to get the software working for the ventilator. Only this week is it ready for FDA testing and approval. Anyone who thinks they can design, code, test, and debug a significant piece of software in less than a weekend is kidding themselves.

[PeterPrinciple]

As a modeler, says Kirkeide, you have complete control over what your results look like.

---

Does she even have a clue to what she has said?

[ThePatriotsFlag]

Most reported models are magnitudes higher than what actually happens. That’s because the news media picks the one with the highest death rate to report. Nobody would read articles saying things are normal. They don’t print news, they print hype.

[SeekAndFind]

RE: Acting in accordance with your beliefs. Nothing to do with truth or science.

One can be sincere but still sincerely wrong.

[rightwingcrazy]

GIGO. If some college kids said “this is our opinion about what’s going to happen, people might ask them for their evidence and reasoning. But if the translate their opinion into a computer model, and generate some dazzling graphs, people instead say “should we write out the check for $2 trillion or $3 trillion?”

[Grampa Dave]

As a modeler, says Kirkeide, you have complete control over what your results look like.

Does she even have a clue to what she has said?

Sounds like what we hear from the Gorebull Warmers like Greta,’It is settled science”!

[McGruff]

Just crank the model up to 11.

[eartick]

These “expert” or no better than the FBI. If they don’t have the outcome they want, they just change the evidence to fit their needs.

[seowulf]

But if the translate their opinion into a computer model, and generate some dazzling graphs, people instead say “should we write out the check for $2 trillion or $3 trillion?”

That's like the power of Powerpoint is to turn bulls*** into gold.

[freeandfreezing]

Anyone who thinks they can design, code, test, and debug a significant piece of software in less than a weekend is kidding themselves.

That is a very important point. Too many college students and other inexperienced developers do not appreciate the complexity of software which does anything significant. They learn to throw together demo level programs largely from existing code and think that they are done since it worked once, or at least it looked like it did.

In the case of the epidemic models many of the modelers fell for the trap of writing a bunch of equations for which you really don't know any of the correct values for the constant terms. You get an output, and if you pick the right values for the parameters it would in fact model what is actually happening in the real world. But you have no idea what the parameters really are, so your model is just a guess.

You can see that in the confidence intervals of the models that have been published. They are so wide that nearly any result falls in their error band.

They are no better than a weather prediction model that says tomorrow's temperature has a greater than 90% chance of being between -30 and + 110 degrees. It is right nearly always, but doesn't tell you much.

[asinclair]

Anyone who thinks they can design, code, test, and debug a significant piece of software in less than a weekend is kidding themselves.

If from scratch, I agree with you. It depends, though, on what you have on which to build the model code. For example, if they had an accurate model of SARS, then it's a matter of fitting the coefficients to the actual numbers you have. Emphasis on the word "accurate."

One step that's not mentioned is how the kids TESTED the accuracy of their model to data to date. Also I'd call into question the accuracy of said data (inaccurate based on reporters says COVID-19 when it may not be, or may not have been the principal cause of death) both as the model was being written, and during subsequent tracking of prediction over reported data.

Questions the reporter did not ask. Must have flunked Statistics 201.

[tbw2]

Neil Ferguson takes the cake. 100x off with the Wuhan coronavirus death rates, 1000x off with mad cow disease and bird flu.

Neil Ferguson, a professor of mathematical biology at Imperial College London, told Guardian Unlimited that up to 200 million people could be killed.
https://www.theguardian.com/world/2005/sep/30/birdflu.jamessturcke

[dila813]

Never use any model that can’t be backtested to at least reference data and the model’s use has to be controlled using only measurements that are available with the reference data.

This is basic stuff. You have to be a complete idiot not to know this.

[LaRueLaDue]

“Never use any model that can’t be backtested to at least reference data and the model’s use has to be controlled using only measurements that are available with the reference data.

This is basic stuff. You have to be a complete idiot not to know this.”

Bingo.

I worked with groundwater flow models and contaminant transport models back in another career, and it could take literally years to get a model working correctly...