Friday links: grant writing advice, skill vs. luck, and more

From Jeremy:

The Contemplative Mammoth has crowd-sourced grant writing advice, aimed at those submitting to the NSF Division of Environmental Biology.

Wired has a nice interview with Michael Mauboussin on untangling skill and luck in sports, investing, and everyday life. Relevant to ecology and evolution too (e.g., think of selection as “skill” and stochastic drift as “luck”). An entertaining, thought-provoking read for academics, and for their undergraduate stats students too. Really drives home the relevance to everyday life of ideas about randomness and sampling.

Did you know there’s a statistical fallacy named after ecologists?! The “ecological fallacy” is the fallacy of assuming that what’s true of the group is true of the subgroups comprising the group. The term was coined by a sociologist in 1950. I say we name a fallacy after sociologists! 😉

Videos of all the panel sessions from the SpotOn London 2012 conference (plugged here a couple of weeks ago) are now on YouTube. Here’s the one from the panel on “the journal is dead, long live the journal”. You can go from that page to watch sessions on all sorts of topics related to online science (broadly defined), from altmetrics to the future of science journalism to data reuse, and much more.

Normal Deviate has a short and provocative post on the differences between frequentist and Bayesian statistics. I basically agree with what he has to say, and I’m reassured that it mostly lines up with things I’ve said in the past. In particular, he astutely notes that “using Bayes’ Theorem” or “using prior information” doesn’t in and of itself make you “Bayesian” (at least not in any way a frequentist would object to). Famed quantitative election forecaster Nate Silver, for instance, sings the praises of Bayes’ Theorem in his new book on prediction, but uses it to achieve frequentist goals. Of course, the post does beg the question of what your scientific goals ought to be, and how that choice dictates your choice of statistical approach (statistics being merely a means to scientific ends). This is an issue Deborah Mayo (for one) has discussed at length. I’ll probably post on some of her ideas in the near future.

From the archives:

Why I don’t care what the biggest question in ecology is. Inspired in part by a great essay by Peter Kareiva on what makes for good ecology. I’m especially curious to hear what my fellow blogger Brian thinks of that essay, since it’s a mix of views I think he agrees with, and views with which he disagrees…

6 thoughts on “Friday links: grant writing advice, skill vs. luck, and more

  1. Interesting link to the Mauboussin interview. I went back and forth in intense debate with two folks at the High Heat Stats blog for over two months this summer on those very topics he discusses, especially the role of “luck”. Much of this originated from the extraordinary success of the Orioles in close games this year, with me taking the stance (at about season’s half way point) that the O’s would continue their success in such games through the remainder of the season, with a number of others there opposing me fiercely on this. [I turned out to be right and now they listen to me a little more closely there]. One of those guys, who said he worked as a consultant to investors, presented some interesting arguments that I’d not heard stated in quite that way before, but now after reading that interview I’m pretty sure where he got his ideas: they’re +/- verbatim Mauboussin’s arguments. We got into essentially every issue he raises in that piece, some of which I agree with him on, and others where he is just flat out wrong in a major way when he applies his concepts to success in baseball (and likely other sports as well). The crux of it is that he is wrong when he says that observations of “reversions to the mean” mean that skill is not a causative factor in that reversion: it very much can be in activities in which human intention plays an important role. He also has a number of undefended assertions in there and uses examples that can cut both ways (for example arguing that the quarterback takes every snap and that therefore football has a higher skill factor than hockey, while completely failing to recognize that a hockey goalie often plays every single minute of a hockey game). And other issues.

    • Yes, thought you might be interested in that interview. I think Mauboussin would freely admit that he’s speculating about why certain sports seem to have a lower skill/luck ratio than others (the quarterback handling the ball so much, vs. the puck being very difficult to keep hold of in hockey, etc.). Anyway, lots of good stuff to chew on in there.

  2. “Science isn’t a body of facts, it’s a body of methods. The most important thing we’ve learned is how to learn.”

    Great statement and I may have to steal it and use it in some lectures. Some of my students get this concept, but for many it’s still about “do I need to know that for the exam?”

    Any thoughts amongst this community as to the best approaches to teaching students how to do science rather than how to “know stuff”? One approach I’ve always favoured is to get them to do as much hands-on, practical problem solving work from the start. But I’d be interested in knowing what others think.

  3. Hi Jeremy – slow to reply to your query as all that turkey in American Thanksgiving made my brain unfit for blogging.

    Can I say both of the above on Peter Kareiva’s essay and Jim’s original essay. In many ways, I find them not as contradictory as Kareiva sets them up to be (PS Jim’s original essay that Peter is responding to is rather hard to find – I had to use URL guessing so it is here)

    I think Jim is right that ecology was less exciting when he wrote that essay (1997) than the 60s/70s. And I do think that ecology got too mired in pulling apart the systems and using the hypotheticodeductive method (teaser for my upcoming blog on prediction in ecology later this week). Its an interesting question whether ecology has found its direction/zing better in the 15 years since he wrote that essay and I would say yes. And I think Jim’s tripartite solution is mostly on target. If I write posts about how scaling up is hard to do I’m certainly not going to argue with his point i) about focusing on emergent phenomena. And I totally resonate with his point iii about needing to evaluate models primarily on empirical success. I wouldn’t necessarily call his approach in point ii) mechanistic like he does, but its hard to argue with the basic formula.

    Kareiva argues that the 60s/70s enthusiasm was misplaced or at least overly optimistic/enthusiastic Hard to argue with this, but it doesn’t mean that the 80s/90s weren’t rather depressing and in need of a shot in the arm. Which is not to say there wasn’t great work being done. The whole emphasis on space and scale that made me who I am as an ecologist burst forth in that era. And I do agree with Peter’s idea that trying to solve some real-world problems would be beneficial to even basic research.

    But to me the bottom line is both authors are grappling with the fact that ecology is more irreducibly complex than, say physics or chemistry. Physics uses “Physics land” simplifications all the time and gets pretty far with it. Ecology tried to use “physics land”-like simplifications in the 60s/70s (which made it exciting) but just didn’t get that far with it. Instead we have had to circle up our wagons and tackle ecology on its terms – many simultaneously non-linearly interacting factors (acting at different scales no less) producing highly contingent systems. The list of factors Peter says were being addressed in the 90s (temporal variability, chaos, ecoevolutionary dynamics, etc) are part of this.

    I find Kareiva’s rejection of Brown’s invocation of medicine as a complex system (“Medicine is an enormously successful reductionist enterprise”) to be a bit of a case of unfortunate timing. While there was a spurt of medical successes that worked on single molecule medicines responding to single protein targets (antibiotics, vaccines, chemotherapy), the truth is most of the things killing people now are the things killing people in the 50s (albeit not the same as the 10s). Things like cancer, heart disease, stroke, dementia. These are enormously complex and not falling to the reductionist paradigm. Indeed I take great satisfaction that after medical biologists lectured us ecologists, essentially calling us a weaker science (See Platt’s strong inference 1964 paper), medical biologists are now embracing the full complexity of systems biology with enthusiasm. Genomics has shown this same trajectory in a compressed time frame. In just a decade we’ve gone from “we’ve sequenced the genome, we’ll soon unlock each gene” to “dang its complicated” (The Mermaid’s Tale blog covers this side of the story well, for example).

    So in summary, I think both describe the same trajectory of ecology from 1960-1997 but through different lens that are both valid. Both identify the fact that ecology is now grappling with innate complexity and see this as a good thing. But on the whole I am more sympathetic with Jim’s solution. Universal laws on emergent phenomena will get us further than continuing to pick apart the details of systems, whether in ecology or medicine.

    • I just want to say well done for guessing a link to the Brown essay to which Kareiva was replying! I have a printed hard copy, and had been toying with the idea of asking Jim for permission to post a scanned pdf of it, just to keep it accessible. I have tried multiple times over the past year to find it via Google or guess the URL.

      Well, that, and thanks for the lengthy reply as well. 😉 The example of medicine is a very interesting one, about which we could probably talk long into the night. For instance, medicine (or at least the large bit concerned with drug discovery) actually is an example of a successful “black box” or “phenomenological” science. You put drugs into people, you put placebos into other people, you record what happens, without worrying about why it happens (i.e. you treat the human body as a “black box”). Even these days, my understanding is that our mechanistic biochemical knowledge of how many drugs work remains rudimentary at best (am I totally off base on that?)

      It won’t surprise you to hear that I’m more sympathetic to Peter’s way of thinking, at least in ecology (I don’t know enough about any other science to have sensible thoughts on how it ought to be pursued…). 😉 But this is one of those judgment calls on which reasonable disagreement is perfectly possible.

      • Good point on drug discovery. Although having limited knowledge, I have a definite sense that for every designer drug that targeted a known protein, we still have multiple medicines that are more or less a random tweak on something mother nature designed and we stumbled upon (ranging from aspirin to tamoxifen). I think our posthoc mechanistic understanding is often pretty good at one level (it binds to this site on this protein) but then it starts to dissolve as we move to a more broad-based understanding (why does aspirin reduce heart attacks?).

        I suspect you and I will have reasonable and friendly disagreement on methodological approaches for the life of this blog 😉 But it makes me keep my arguments sharp knowing you will be commenting on it! And in the end diversity of approaches to ecology is good for ecology. A single scientific method is a myth.

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 )

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.