Models: Garbage In, Garbage Out

I watched Governor Walz’s presser last night. My impressions (borrowing a bit from David Strom):

Old And In The Way – The data used in this model is 3 weeks old. The pandemic is three months old in Minnesota at this point. That means the model is ignoring a solid quarter, almost a third, of the data available – and, being most recent, very possibly the best data. I hate to throw the word “useless” around willy-nilly – but if this alone doesn’t make the model useless, I’m completely bumfuzzled.

Older And Out Of The Way – The model don’t distinguish between populations in hotspots like Hennepin County and the rest of the state. In Minnesota, at present,99% of the Covid deaths are concentrated in 3% of the population. 80+% of the deaths involve 1% of the state. And yet they apply those percentages to the entire population. This is the sort of thing that’d get sent back for rework in the private sector.

And basing mortality predictions across the entire population on numbers that are so flawed at the concept level is bad math at best, dishonest at worst, and a tool used to deceive the people either way – the equivalent of a Star-Tribune poll that draws statewide conclusions based on a massive oversampling of Minneapolis DFLers.

Under Bus Shoved – And since the overwhelming majority of deaths involve people in long term care, the model tacitly assumes that nursing home residents will do nothing whatsoever to protect their population. Which is possible, although there are a whoooole lot of lawyers out there who will likely impel them to try to do a lot better than they are, sooner than later.

The model banks on long-term incompetence – natural, perhaps, for bureaucrats, but not necessarily representative of the population at large.

Conclusion – the Walz administration is using this model to flim-flam a state that, signs show, is getting tired of being lied, condescended and talked down to.

10 thoughts on “Models: Garbage In, Garbage Out

  1. Minnesota numbers not mentioned in the model:

    Not one child has died from it.
    Not one teen has died from it.
    Not one 20-something has died from it.

    2 people under 40, and 6 people under 50, have died from it.

    If you are under 60, your odds of dying from it are 40 in 5,500,000. That’s 1 in 137,000. You’re more likely to get hit by lightning, die from a bee sting, or sink a hole-in-one.

    If you’re 60 or older, or if you know you have risk factors, then act as if you were a grown-up adult possessing intelligence guided by experience. Take steps to protect yourself.

    Everybody else – you have basically zero risk. Get to work.

  2. The Great and Powerful OZ had his curtain to hide his sham….Walz has his own curtain and it’s being exposed. Am I the only one that thought he seemed to be in quite a huff while delivering his message last night? The guy actually seemed pissed to me.

  3. The least safe people are the ones who, ironically, are the most “stay-at-home”: the nursing home population. As JD pointed out (and as MN’s own numbers from 5/11 show), there are a handful of deaths under age 60. The risk is low, and the question for the nursing homes (which are overseen by the State), why is the death-toll climbing? The ill aren’t playing canasta and doing jello-shots with each other in these places, and yet it’s spreading.

    That said, I have two friends under age 60, with no co-morbidities I’m aware of, who caught the virus and it was no fun. Literally weeks of feeling desperately ill, weak and miserable. This is nothing to dismiss as an inconvenience. So be wise and look out for yourself and for others as you go about. Yes, the overwhelming number of people who get it suffer mild symptoms, but I don’t think anyone knows (or is sharing) why some healthy people suffer more than others. So be wise. To borrow something Joel Rosenberg used to say, “Test cases are for other people.”

  4. It is striking me more and more that what we may need to do is ask those who are not as vulnerable to go to “Camp Covid” and deliberately get the disease to advance to herd immunity. That will be a tough sell, since those who can’t affect the vulnerable are generally the “single”, and they have a short term time preference.

  5. Reason that the disease is spreading in nursing homes is because of who works there–I’m the dad of three beautiful young ladies who work at a local nursing home, and they comment that many of their coworkers are prone to showing up a touch late, hung over, etc.. It frustrates the heck out of them because you can’t just find CNAs everywhere (cleaning out bedpans is surprisingly not a high demand job), so you can’t just fire people on the first or even the tenth offense. So you get a number of people who just don’t totally care, and I’d bet a nickel they’re the primary vector for the disease into nursing homes.

  6. Words of Wisdom from Walz:
    “This is either going to work or not work. People are either going to stay out of the hospital or get in it.”
    Now that is leadership, Minnesota style.

  7. More than 80 percent of those who’ve died from the disease in Minnesota were living in long-term care facilities, with most having some underlying health problems.
    Still, the department statistics noted a third person in their 30s had died from COVID-19.
    No curiosity about this “person in their 30s”? Note misuse of the plural pronoun to refer to an individual. How dehumanizing.
    In at least one case in Hawaii, the relatively young victim (in his 50s) was a homosexual compromised by an STD. It took a long time for that info to be made public, out of “privacy concerns.”

  8. Pingback: Kung Flu in MN | Things to Remember, Things to Cherish

  9. Models are just that: models. And how good they are ultimately depends on how well they work, and if what they do generates value.

    These models were written by academics and were written so academics could write papers on what a pandemic might look like. They met that purpose but are now being relied upon to guide policy in the real world, despite having never been tested that way, nor having had adequate review.

    I compare that to the tools I use. Sure, the circuit simulator SPICE was written by a professor long, long ago. But he was an EE, understood models, understood finite element mathematics, understood how errors could propagate through systems and took care to get it as right as you could in those days. As others took over they had very, very large financial risks and rewards to get their circuit simulators right, and they got constant feedback on what worked and what didn’t. These days the models and simulations aren’t perfect, but they’re good enough that companies are willing to multi-billion dollar projects on their results.

    I’ve said in conjunction with the climate models, and I’ll repeat it in conjunction with these models: If semiconductor modelling were as good as epidemiology modelling, we’d still be using abacuses rather than computers.

  10. Pingback: In The Mailbox: 05.15.20 (Afternoon Edition) : The Other McCain

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.