Friday links: Blackboard vs. baby clothes, Bayesian vs. Bayesian, and more

Also this week: teaching scientific writing using silly social media fights, Industrial Revolution counterfactuals, Jensen’s inequality vs. dark matter, celebrating (?) tenure, and more.

From Jeremy:

Dan Simpson with an interesting post on why he hates the “8 schools” example often used to teach Bayesian hierarchical models. Even if you don’t care about teaching Bayesian hierarchical models, you should read this for Dan’s provocative thoughts on why simulated data are better than real data for evaluating the performance of any statistical approach. I also found this interesting because it’s a case of disagreement between two bloggers who agree on a lot. Dan Simpson blogs at Andrew Gelman’s blog, and Gelman likes the “8 schools” example. Now I’m wondering if there’s any ecological or academic topic on which I seriously disagree with Brian and/or Meghan, and if so, what it is and how to identify it. Brian and I sometimes think we seriously disagree, but after a bit of discussion generally find that we mostly agree.

How Blackboard is like baby clothes. (ht Marginal Revolution)

A toy model of scientific research in which there’s selection for scientists who do low-risk, low-reward research because (i) risk-taking researchers will experience more failures and so will train fewer members of the next generation of scientists, and (ii) it’s hard to train students to do high-risk, high-reward science. Fun to think about, but just offhand I’m not at all convinced it’s a good caricature of reality.

Why did the Industrial Revolution first take off in Britain rather than France?

Psychology papers mostly use Bayes factors as an excuse to infer the truth of the null hypothesis, in cases where sample sizes are small and the data are noisy. I don’t feel like Bayes factors are much used in ecology, but I’d be curious to know if they’re used in the same way when they are used. Anyone care to check?

Are dark matter and dark energy actually just artifacts of ignoring Jensen’s inequality?

Explaining evolution to non-scientists: lessons from Ross and Phoebe. Irrelevant aside: I wish we lived in a world in which someone could feel free to write this post without first acknowledging critiques of Friends that have nothing to do with the particular scene on which the post is based.

I’m months late to this celebration of getting tenure, which I hope and assume is entirely in jest.

Coleen Rooney demonstrates how to structure a scientific paper. 🙂 I assume that Stephen Heard will incorporate this example into the second edition of his scientific writing book.

As a former extremely slow distance runner, I just want to say OMG. Twice.

I wonder if I can commission a statue to commemorate my discovery of alien life? 🙂

7 thoughts on “Friday links: Blackboard vs. baby clothes, Bayesian vs. Bayesian, and more

  1. I think the toy model of science got the pattern (a favoring of low risk, low-reward science) right but the wrong reasons. I don’t think this goes through students. Students are if anything more likely to do high risk science (less to risk for one thing). I think grants and the increasing trend of assessing our peers by quantity rather than quality (due in turn to the Red Queen acceleration of productivity demands) have more to do with it.

    I *LOVE* the dark matter/Jensen’s inequality post. Could it be that the Jensen’s inequality correction terms are small on a solar system scale but large on a galactic scale? I find it kind of mind boggling that the whole field has resorted to fudge factors rather than assess the magnitude of error. I get that assessment of error is not easy, but couldn’t some simulations get at that these days? But then I have written here in the blog and in a paper that Jensen’s inequality in the context nonlinearity and variance (very common in ecology) are at the heart of why microprocesses fail to easily scale to macroprocesses (and Robert O’Neill got there long before I did). It seems the physics example is not that dissimilar. Nonlinear gravity/space curvature equations and heterogeneous irregular density of mass (Einstein collapses to Newton in the solar system scale only because of symmetries of nearly spherical planets)

    • Yeah, I was really intrigued by that post too. I almost wish I was a highly-trained astrophysicist just so that I could have a more informed opinion. On the one hand, there certainly are plenty of cases in which smart, experienced people can just eyeball a nonlinear correction factor (or higher-order term in a Taylor approximation, or etc.), go “eh, that’ll be too small to be worth calculating”, and be right. But on the other hand, the only reason to think that dark matter and dark energy exist is that a linear prediction breaks down at galactic scales. So it’s *really* important to make sure that that linear prediction is actually a sufficiently-good approximation to the underlying nonlinear reality! And given that we’ve been trying and failing to find dark matter and dark energy for decades, maybe it’s time to go back to the drawing board and double check whether that nonlinear correction factor really is as small as most people seem to think it is. As you (and Sabine Hossenfelder, and some of her commenters) say, it shouldn’t be *that* hard to simulate a toy model of the universe and see if the nonlinear correction factor is small in that toy model.

    • And yeah, agree that low risk, low reward science isn’t favored primarily because students who do high risk projects are disproportionately likely to have to drop out of science.

    • It may be worth pointing out that the only substantive discussion of this issue which Sabine Hossenfelder pointed to (the link at the end of her post) is about whether averaging over the entire universe affects general cosmological calculations of the evolution of the universe; the speculation is that this might, perhaps, explain the accelerating expansion of the universe that’s normally explained as the effect of “dark energy”.

      This has, as far as I can tell, nothing at all to do with questions of dark matter within galaxies.

  2. Einstein collapses to Newton in the solar system scale only because of symmetries of nearly spherical planets

    No, Einstein collapses to Newton in the limit of low energies, which is something Einstein worked out when he first devised general relativity. It has nothing to do with symmetries, spherical or otherwise (and regardless of how spherical individual planets may or may not be, the solar system is not spherically symmetric).

    (A non-rotating black hole has perfect spherical symmetry, but it does not follow Newtonian physics.)

  3. It’s probably not right to think there’s high risk, high reward science, and low risk, low reward science.

    The last time I was reviewing grants, NASA wanted us to do some exercise – I forget exactly how it was structured – to identify high risk, high reward proposals so they’re not getting overlooked. The panel was pretty unanimous we were funding less than half of the low risk, high reward proposals, so of course we weren’t funding any high risk, high reward stuff (and certainly not any low risk, low reward stuff).

    And not to be a buzzkill, but globular clusters substantially overlap with dwarf galaxies in baryonic mass, which makes it pretty clear there really is dark matter*. Dark energy – I’m far less comfortable making any kind of confident statements.

    *Jeez, I spent so much time trying to make a Bullet Cluster joke here, which also makes it very clear there really is dark matter. But I couldn’t make it work.

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

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