8 thoughts on “Can We Trust Covid Modeling?”

  1. Too many conclusions from too little information. Starting to look like we should follow the money and take it back from those who are collecting it.

  2. Sweden has low pop density and it’s surrounded by countries in lockdown.
    South America is better test- but it’s death per million is still low.

    Part of lockdown of course is stopping international air travel- And US doing that is like a global lockdown.
    But part of US lockdown is to prevent more deaths in rest of World- something China didn’t do.

  3. I dunno. How many prior outbreaks were tracked carefully enough to supply enough real world data to compare real world results to the model runs? Like H1N1 or SARS? Or even the “Spanish flu” outbreak of 1918? How well did they perform against the historical data? It’s perfectly acceptable to modify parameters if you can get a good curve fit to the historical data and then plug in current parameters as best they are known. It’s not ok to just make something up for SARS-CoV2 based on what you GUESS are best fit parameters or to curve fit after the fact. After the fact might be fine for SARS-CoV3 but not this outbreak., It is also NOT OK to average results from a stochastic model that shows non-numerical random behavior. Anyone bother? Or are we just into scifi? I am hopeful. There are, after all, many other models besides Ferguson20. But then again, there aren’t any FDA trials and approvals needed to avoid model quackery. Maybe, just on a wild hunch, there should be? Pass me that slice of baloney that looks like a finger and that model result that looks like a roulette wheel.

  4. I have frequently pointed out to people the conundrum of predicting avoidable disasters. If you convince people that a particular airline flight is going to crash and kill 200 people, it can’t crash and kill 200 people because nobody will buy a ticket.

    For a prediction of an avoidable catastrophe to be accurate, it has to be ignored, in which case it’s useless for anything other than winning bar bets. If it’s useful, it also turns out to be completely wrong.

    It’s kind of like a sci-fi time travel paradox.

  5. It was a set of computer models that prompted politicians around the world to destroy their own economies. Those models predicted deaths in the millions. That has not happened, not even close. How much of that difference can be attributed to people changing their behavior and social distancing and how much is due to the models being wrong may never be known for sure. We don’t have the benefit of a time machine that allows us to go back and run alternate scenarios.

    Given the multi-trillion dollar impact just in the US of these model predictions, I think we need to set up some quality and certification standards for computer models that will be used to guide public policy. Some ideas that come to mind include:
    1. Open source of the model computer code to allow independent examination.
    2. Open source of the model input data.
    3. An approved set of qualification criteria, to include things like consistency of outputs for a given input, sensitivity of the model to small changes in input data, and other criteria to certify the model is up to current professional standards.
    4. Version control to carefully manage all updates to ensure nothing gets broken.
    5. Where applicable, continuous testing to validate that the predictions match reality. If you model doesn’t agree with the real world, it isn’t the real world that’s wrong.

    None of these ideas are new. If you work for an aviation avionics company, the FAA has processes in place to regulate software quality and reliability. Companies like Garmin can’t just change their source code and pump it out to their aviation customers. This does slow the rate of change, but it slows the rate of errors, too.

    1. –Given the multi-trillion dollar impact just in the US of these model predictions, I think we need to set up some quality and certification standards for computer models that will be used to guide public policy.–

      There is no model that can predict the future.
      As IPCC says no climate model can predict the future.
      And I think that is they only they said which is correct.
      What all model are is projections about what someone thinks
      is going to happen in the future.
      One thing about models is if computer gives a result which doesn’t
      fit what a person thinks is going to happen in the future, they change how something is modeled, until it fits.
      And all people tend to want a model to exaggerate, which tends to “go higher” then they imagine it will be, but have error bar or low estimate where think it’s going to be in “worse” case.
      Or tend to want the audience to imagine it’s going to higher than They imagine could “possibly be”. Because they selling something, and the low range tends to closer to what their fevered brains imagine is true {though on the high side on what imagine will happen}.
      And this roughly fits what normal people do- they expect something, which they could predict fairly well, to have the possibility to something different. They look down the street a see car coming, they predict it’s speed, and allow the car may going faster than expected. It’s not safe to guess car is going slower.
      And certain not safe {or profitable} for modeler to under estimate.
      I never encountered anyone who understand global climate.
      And no one understands this virus.
      Simple things like wearing a mask or washing hands- I don’t “know” it helps. But I will wear a mask and wash my hands- because it’s relatively simple to do, and it might be a significant factor.
      But I tend to think this virus is airborne spread- though that allows it to be on surface, which can be touched, and transmitted by touch to face.
      Anyways, I under predicted the effects of this virus, But I was allowing for lockdown to be effective. And I knew and said I was not over estimating it. And said if was in position of Authority, I would not give such low estimate- I would give higher estimate.
      They quite reasons I would give a higher estimate if I was in position of Authority. A primary reason is there was a lot known about the virus. And at the moment there is a lot of unknowns.

      In terms of general thing about “lockdown measures” there some things I feel were correct:
      Stopping air travel from China, I am 99% sure was correct.
      Or 99% certain it worth the costs to do it, and reduced a lot deaths in US, and more significantly, reduced deaths, globally.
      And if stopped air travel from China, sooner, it would be better.
      And if stopped 1 week later, it might have had far worse consequences. And more certain if stopped a week earlier than US did, it would have been far better.
      Or taking unfair advantage of hindsight, it seems easy to say stopping it earlier, could have had huge difference. Or it seems if Trump delayed doing this, it would have been a huge mistake.
      So in terms of lockdown measure, Trump gets at least a B+ in terms that action taken. But if want to be more generous, A+ would not be unreasonable- even though shutting earlier would be have better.

      1. Let’s not forget how any attempt by Trump to do a lock down in January (even February) would have been perceived by the MSM during the height of the impeachment process.

Comments are closed.