Pages

Tuesday, March 17, 2020

You can't trust the weather man

1. Thanks to Steve Hays for sending me this article by Dr. John Ioannidis. Ioannidis is a Greek physician and epidemiologist. He's most famous for (rightly) taking to task academic scientific publications and showing how the vast majority of them are basically bad science and/or bad statistics. Ioannidis' article on the coronavirus is worth reading. As an ID physician said about the same article: "Excellent piece by ⁦@METRICStanford on CFRs, the calculus of social distancing, and how lack of data is compromising decisions."

2. Similarly I earlier pointed out that Nate Silver (whose day job is predicting political elections on his 538 website) brought up some limitations in coronavirus models too.

3. I think epidemiological models might parallel meteorological models. That is, I think epidemiological models attempting to predict how bad the coronavirus will be might be similar to weather reports trying to predict the weather. We can't accurately predict the weather 100% of the time. Let alone in every place in the world. Sure, there may be general accuracy, but weather reports notoriously get things wrong too. Just my general impression, but I'm no epidemiologist or meteorologist or statistician or the like.

Of course, this doesn't necessarily mean we shouldn't bring an umbrella if the weather report predicts it's going to rain, but it also doesn't necessarily mean we should. It depends on a number of factors. Such as the further out the weather forecast is, the more unreliable it is likely to be. It's normally more accurate to predict it will rain this evening than it is to predict it will rain tomorrow or one week from now or one month from now.

4. I guess the only sure thing about predictions regarding the weather, political elections, the future impact of disease on populations, what the stock market will do, and who will win the Super Bowl is that there's bound to be someone in the end who will say: "See, I told you so!" :) And sometimes they might even be right. Sometimes.

8 comments:

  1. Whatever happens regarding this virus, I'm not sure how we can ever go back to a time when "merely" 670,000 hospitalizations and 50,000 or so deaths over six months don't merit similar extreme measures.

    What, those 50,000 people don't matter? What's the threshold for acceptable and not acceptable deaths?

    What happens when we start to see more resistant bacteria or yet another strain of these viruses? Wash, rinse, repeat?

    https://www.cdc.gov/flu/about/burden/preliminary-in-season-estimates.htm

    I'm going to give our government leaders the benefit of the doubt and assume that there is a built-in expectation of needed care and that this virus was potentially going to exceed that by far. Hence, the panic.

    If that is the case, then it seems like we need to find a way to enable our health care system to be able to respond more dynamically to emerging health care crises (such as pandemics like this). In practice, I think this means diversifying our sources for equipment and pharmaceuticals and not relying on single sources (or countries, particularly potentially hostile ones).

    ReplyDelete
    Replies
    1. "In practice, I think this means diversifying our sources for equipment and pharmaceuticals and not relying on single sources (or countries, particularly potentially hostile ones)."

      Good point! I agree.

      Delete
    2. "we need to find a way to enable our health care system to be able to respond more dynamically to emerging health care crises (such as pandemics like this)."

      For solutions, you might be interested in Michael Osterholm's work. Such as his book Deadliest Enemy. I think the last chapter offers some ideas for what we can do. It was published in 2017 before this pandemic. Experts like Osterholm have been warning about these things for a long time. That's not to suggest I agree with his solutions, but at least he's worth hearing out.

      Delete
  2. In my beloved Denver you might as well throw bones as trust the weather man. I don’t know if that fits this post I just felt like saying it.

    ReplyDelete
    Replies
    1. Lol, I like the saying, so I'm going to go with a big thumb's up. :)

      Delete
  3. Three things although I am no longer in the sciences:

    1) The models will always be problematic but we have no choice but to pursue them. The BBC
    pointed out their shortcomings.

    2) A lot of science journals and stuff are fluff. However when they are good, they are really good.

    3) That aside - on an interesting and more positive note:

    A lot of pollution is being cleaned up on account of the virus. For the first time in ages, the waters of Venice are starting to run clean.

    ReplyDelete
    Replies
    1. Thanks, rgbrao! :)

      Oh yeah, I think I pointed out something similar in one of my recent posts (here). The models are significantly flawed, but I'm not sure if we can do much better without them.

      Very good point and very true about the good science papers. There are many that are classic. Depending on one's field, of course.

      I've heard the same about the Venetian canals. Also, you might have already seen this, but for those who haven't, apparently dolphins have started to come closer to ports again since ferries aren't going out as much anymore due to the coronavirus.

      Delete
    2. Sorry, it looks like they removed the tweet with the dolphin! I only saw it yesterday or maybe the day before.

      Delete