Login
You're viewing the mastodonapp.uk public feed.
  • Daniel Lakelanddlakelan@mastodon.sdf.org
    Oct 31, 2026, 1:38 AM

    @futurebird

    I think when teaching statistics, the first place to start is with a set of experiments you are going to run. Literal physical experiments. For example you could drop little paper helicopters with a paperclip and measure their flight time. You could drop balls of crumpled paper and measure their drop time. You could roll balls down ramps and measure the rolling time. You could measure the time to drain a funnel with different types of cooking oil, at different temperatures.

    💬 3🔄 6⭐ 0

Replies

  • Daniel Lakelanddlakelan@mastodon.sdf.org
    Oct 31, 2026, 1:42 AM

    @futurebird

    Whatever these experiments are, you should be talking about the instrumentation you're using to do the measurements, and the ways in which your measurements might not fully accurately reflect the reality of what happened.

    I did an experiment where I took a plastic disposable pipette and dripped water onto a scale, and did so after cutting the pipette shorter and shorter and measuring the diameter of the opening each time... And my kids and I built a model of surface tension

    💬 1🔄 0⭐ 0
  • Daniel Lakelanddlakelan@mastodon.sdf.org
    Oct 31, 2026, 1:43 AM

    @futurebird

    So, this is why I say math is not really the key in a good stats textbook, because what really matters is thinking about connecting an experiment to a model, and you can do pretty easy models that require only basic algebra and things.

    💬 2🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Daniel Lakelanddlakelan@mastodon.sdf.org
    Oct 31, 2026, 1:45 AM

    @futurebird

    The surface tension model was great because most scales have a nonlinear behavior near zero! So now you're modeling both the droplet formation and the forces from surface adhesion, and the measurement scale and its nonlinearity for small mass/weight. It doesn't take more than algebra really but it does take thinking clearly about which particular algebra!

    💬 1🔄 0⭐ 0
  • David Chisnall (*Now with 50% more sarcasm!*)david_chisnall@infosec.exchange
    Oct 31, 2026, 7:39 AM

    @dlakelan @futurebird

    100% agreed. I always found statistics to be a bunch of weird maths that did… something? But not for reasons anyone could explain to me, just memorise the formula.

    Then someone explained to me that statistics is about determining the likelihood that a model corresponds to reality. So you need to have a model and you need to have some experiment, and you’re trying to make sure that the experiment samples enough of the space that each new sample can (ideally) eliminate some alternative models and give confidence to your model.

    And, once I understood that, a bunch of statistical techniques made sense because they’re just trivial applications of probability. And those weird formulae are just reusable partial applications, like the quadratic equation formula that is just the result of completing the square on a quadratic where all of the variables are symbolic.

    💬 1🔄 0⭐ 0
  • Daniel Lakelanddlakelan@mastodon.sdf.org
    Oct 31, 2026, 8:55 AM

    @david_chisnall
    I love this so much. Yes, the formulas for quantiles of the standard normal and all that stuff is just that, formulas, by themselves they have no meaning, theyre just math facts like 2+2=4 but the whole point of statistics is "how well does this model correspond to reality?"
    @futurebird

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Three plus or minus fiveThreeSigma@mastodon.online
    Oct 31, 2026, 3:30 AM

    @futurebird @dlakelan
    Although I learned formally, my -understanding- of stats and human overconfidence comes from an experiment in grad school, working with another small group watching data points slowly slowly appear in our spectrograph. We weren’t sure if there would be peaks showing up but we hoped so. So we were oils say things like “I bet that’s the 8Li peak” and slowly get disappointed as our bets turned out wrong.

    💬 0🔄 0⭐ 0
  • ubiubi@ecoevo.social
    Oct 31, 2026, 4:31 AM

    @dlakelan @futurebird When teaching stats I like to give a physical example of a parameter (eg a die) that gives rise to a random observation (a random number). And I repeat myself many times that in science we are often trying to study a hidden unknown parameter, using only the random observations that it generates.

    💬 0🔄 0⭐ 0