Friday links: regression to the mean vs. the purple-snouted crompus, profs vs. meetings, and more

Also this week: fighting over frequency, the challenges of writing public-facing scholarship, citizen squirrel science, T. rex gets a PhD, and more.

From Jeremy:

Stephen Heard with a blog post about a very nice and important paper on which he’s a co-author, that grew out of a previous blog post of his. The previous post suggested, and the subsequent paper shows, that many species that have been monitored over time may appear to be declining in abundance not because they actually are declining, but because of regression to the mean. If species’ abundances vary in time and space (without any long-term trend in time), but you only start monitoring them at the times and places where they’re most abundant, you’re going to tend to see declines in abundance. Nice example of using a blog post to put an interesting, provocative idea out there quickly, get feedback on it, and then use that feedback to guide further research that firms up the idea. I’ve done the same thing myself a couple of times (example), with a third in press. Will have to add Stephen’s paper to my list of statistical vignettes for teaching intro biostats. Also, +1 to the paper’s authors for working in “purple-snouted crompus”. 🙂 Now I want every ecology paper that involves a hypothetical example to give that example a ridiculous name. I claim “Mortimer’s Snerd” for my next hypothetical species. Well, unless I need a hypothetical example of a declining or extinct species, in which case I claim “ambiguous pazuma“.

Artem Kaznatcheev reports on what sounds like a very interesting talk by philosopher of science Karen Kovaka. Who observes that heated scientific debates over a seemingly straightforward empirical question–“How frequent is X?”–generally don’t get resolved. Usually they just peter out. Now I really, really want to see the paper when it comes out so that I can dive into this.

A takedown of how sexual harassment is (not) addressed in the workplace. Focuses on corporate workplaces but applies to academia too. (ht Brian)

Political scientist and FiveThirtyEight writer Julia Azari on the challenges of writing “public-facing scholarship” on venues like Twitter and FiveThirtyEight. Particularly when you–and the publication for which you’re writing–openly adopt a particular political point of view. Curious to hear if this resonates with folks who write about ecological research in a public-facing way. I’m thinking for instance of ecologists’ public-facing writings about the recent “Insectageddon” controversy. (ht @jtlevy)

Vaguely related to the previous: what, if anything, should (or can) researchers do about the fact that, when they tweet out a brief summary of a paper (often accompanied by a single figure, or a screenshot of the abstract or conclusions), hardly anybody clicks through and reads the actual paper? Is that a good thing on balance? After all, it’s probably a few more people than would’ve read the paper otherwise. Or is it a bad thing on balance? Because if a little knowledge is a dangerous thing, the tweet probably greatly increased the number of people who only have a little knowledge. And yes, one absolutely could ask the same questions about our linkfest entries…

Attention Stephen Heard: citizen scientists see squirrels. 🙂 This is interesting as well as amusing. The project seems to have started as a joke, and the founder still emphasizes that it’s a way of “telling a story” rather than a scientific study. But yet, the project volunteers seem to have put a lot of effort into collecting detailed, high-quality data, and the data will be publicly released. So, is it science or not? Or is that the wrong question to ask? (ht Matt Levine)

And finally: T. rex gets a PhD. 😛 😛 😛

From Meghan:

LOLsob at this McSweeney’s advice on how to chair an academic committee. (More seriously: there are resources about how to run meetings, set agendas, etc. If you will be chairing a committee or running meetings, please read them!) (Jeremy adds: Brian spent a decade in the business world. He knows how to run a meeting. Here’s his post on how to do it.)

20 thoughts on “Friday links: regression to the mean vs. the purple-snouted crompus, profs vs. meetings, and more

  1. Glad you liked “purple-snouted crompus”. I had to fight tooth and nail to keep that, against a copyeditor who thought it wasn’t appropriate (and who AFTER we restored it in the copyedit phase, STILL cut the Latin “Crompus nasopurpurea” again in preparing proofs). (No, you’re right, that kind of thing is not what professional copyeditors do…)

    I’m pleased to have this post as a concrete example of how a little bit of whimsy can engage people with writing. Thanks!

    • My own experience with copyeditors at journals has been mixed. I’m sure there are many excellent copyeditors in the world, and I would never bash them as a profession. But in my own anecdotal experience over the years, the copyeditors on my papers have been a very mixed bag. On many of my papers, they’ve introduced at least one grammatical mistake, and tried to impose at least one “clarification” that actually changed my meaning. And as you say, it’s not their place to remove any text (even humorous text) that’s been through peer review, and *especially* not that’s been through the proof stage.

    • Something I hear rarely talked about is how often copyeditors these days are “off-shored”. With ConBio being a Wiley journal it was very likely copyedited by somebody in the Phillipines or Singapore or such. That applies to most journals. Modern copy editing services don’t hold a candle to the real high copy editing that used to occur 20 years ago. The handful of times I have gotten a real high quality copy editor (e.g AmNat, AJB) it just jumps out at you how wonderful it can to have a professional backstopping you.

      Whereas I agree the common experience these days is to have them introduce as many problems as they fix. And to be rather cavalier in what their role is.

  2. On regression to the mean: In North America, forest cover had been greatly diminished by the 1920s (just about 100 years ago) from exploitative logging and clearing for agriculture. Back then, most (?) people in the US lived on family farms that combined vegetable crops, pasture, and hayfields for the horses that powered everything. As people left farms for cities and suburbs in the 20th Century – and with horses rapidly replaced by tractors – a huge amount of that open land succeeded to shrubs and back to forest. Today, we actually have pretty close to what we think might have been overall forest cover in North America at the dawn of European colonization (though we’re learning all the time that indigenous people might have had more of the land cleared for their agriculture than generally appreciated). One could get whiplash from looking at forest cover 1800–>1900–>2000.

    How does this color our perceptions? Population estimates for North American birds are, for most species, considered with respect to a baseline from our most rigorous and comprehensive monitoring program, the North American Breeding Bird Survey. The BBS began in 1966. This makes the baseline something like data from 1966–1976, i.e., mid- to late-20th Century during the time when small family farms were being rapidly abandoned but there was still a lot more pasture and hayfield distributed around the country. This means that what we think about baseline bird populations – especially grassland birds – was probably an elevated artifact of extreme human landscape modification in the roughly 200 years prior.

    To be clear, I am *absolutely convinced that our grassland birds are in trouble in the US and in dire need of conservation action*. They’re a big part of my research program, too. That’s mostly due to the fact that the open lands we do have today and the way we manage them are largely incompatible with the needs of grassland birds. But statements such as “this species has declined by ___ %” are likely misleading and incomparable among species because of . . . regression to the mean.

    • A question re: regression to the mean that just occurred to me and that I’m still mulling over. Is there such a thing as regression to the mean in nonstationary systems that lack a well-defined mean?

      I was prompted to think about this by your remark that, because of a unique confluence of historical circumstances that are unlikely to ever be repeated, abundances of US grassland birds may well have been high in the 1960s, relative to historical norms. But what if there’s no such thing as a “historical norm”? What if history is just one damned thing after another, as the old saying goes? What if one or more of the factors that affect bird abundances do an unbounded random walk, so that bird abundances do too (at least if you monitor them for long enough)?

      I’m far from the first ecologist to have this thought, of course. Curious to hear what others make of it in the context of regression to the mean.

      • I wouldn’t consider “starting at a historical high” regression to the mean. But the two combine: if we have 1000 fields, and count is generated as a random draw from the historical mean curve for the region (the mean of all the fields over time) – so the residual from the curve is entirely random, and we start our study, at any point in history, on a field that has a count well above this historical mean curve, then the expectation is that future samples from this field would be nearer the historical mean curve.

      • Yes this issue of appropriate baseline is a major issue for any attempt to define loss. Many people are convinced that an analysis of trends in biodiversity are valid only if they start in the appropriate time period. But between climate and humans there is no obvious appropriate time. Change is everpresent. And if one wants a “pre-human” baseline one really needs to go back at least 10,000 years which nobody does.

      • Agree 100% with all of this. I’m mystified whenever somebody complains that failure to find declines in diversity or abundance *must* mean that the time series data started at the “wrong” time, or at some “weird” time.

      • “But what if there’s no such thing as a “historical norm”? What if history is just one damned thing after another, as the old saying goes?…”

        ¡Exactamente, señor! ¡Muchas gracias! 🙂

        i.e., there is no “equilibrium” baseline: a given set of organisms and relationships is an open system, analogous to a hypersaline pond getting splashed one day with normal sea water + organisms, the next day with an infusion of freshwater + organisms, the next attacked by a flock of pelicans, and the next day drained by a break in the reef, the next spoiled by an undersea vent release….; one event after another so no stable equilibrium is ever reached.

        The assumption of equilibrium is useful for illustrative models but I suspect frequently wrong in reality.

      • I think you may have misunderstood me Jim. My question has nothing to do with equilibrium or the lack thereof. Stationarity is a much broader concept–it just means that the statistical distribution of the variable of interest doesn’t change over time. The productivity (or whatever) of that hypothetical hypersaline pond could bounce around radically from one day to the next due to all sorts of perturbations. But so long as one chunk of the time series of productivity measurements looks like any other chunk (same mean, variance, etc.), it’s stationary. In a stationary world, the concept of regression to the mean makes perfect sense. Because the concept of “typical conditions” (or “typical distribution of conditions) makes perfect sense.

    • “I think you may have misunderstood me…”

      thanks, J. Thanks not inconceivable…. 🙂

      but I’ll take a crack at clarifying myself:

      I referred to “dynamic equilibrium” to express a system that regularly experiences departure from and regression to the mean. I presumed from what you said that a system varies about the mean but a stationary or stable system returns to the mean, presumably because there are no processes that create a large enough perturbations to drive it off the mean on the time frame of the observations. So, one observer might see a certain population level of a species, and a later observer might see a higher population level, but not recognize that the variation is “normal” variation about the mean, but on a somewhat larger timescale than the observer’s time scale. I suspect that such systems are less common than might be believed, because interactions can occur on a huge variety of time scales and might not be perceptible to the observer.

      The way I understand what you’ve said is that a system might vary around the mean, then experience a large perturbation, start to return to the mean, but before it can do so it experiences another large scale even that drives it in yet a different direction. This is what I understand as “history is just one damned thing after another” – the system can’t return to a mean because the relationships are continually being perturbed in different a new directions that equilibrate on different time scales.

      The hypersaline pond was intended to show an environment and contained relationships that never achieve stability because each “day” brings new, unpredictable events that shift the system in new ways. The time frame could be months, years, decades or centuries etc, but it seems like that would depend on the lifespan of the inhabitants and the frequency and magnitude of the perturbations.

      Perhaps I’m still missing the point but at least that’s a better explanation of what’s in my head. thanks!

  3. Right, there’s nothing that anything is “supposed” to be, and no ecological justification for the selection of a baseline for restoration. Our baselines are arbitrary, and that’s okay. It is good, however, to have thought about our choices.

    I’ve had to spend a lot of time considering this in the development of ecological indicators for broad-scale assessments. My (imperfect) default is to consider the spatial scale over which the assessment is desired and examine the species that evolved within/encountered conditions to which they were adapted in that area. Can populations of those species be maintained in the current landscape or within bounds of projected landscape changes? I’m looking for the opportunity for species that evolved in a region to continue to do so. Thus it doesn’t matter to me if, for example, Northern Bobwhite was *anthropogenically* abundant in the 1960s when it’s North American population baseline was established. I’m interested in the more coarsely grained “is this species native here?” at all. If so, and it’s not doing well, then conservation action is justified regardless of the population baseline.

    • I remember learning in grad school that, for restoration, the appropriate baseline is when you were a child. This was very much tongue-in-cheek, and a good way of starting a discussion on the difficulties of defining the baseline.

    • 🙂

      I write population ecology exam questions about jackalopes.

      Basing homework and exam questions on made-up species actually serves a serious pedagogical purpose. If I want students to focus on answering the question asked rather than giving me irrelevant information, I base the question on a made-up species. That way the students can’t give me irrelevant information, because the only information they have is what I provided in the question.

  4. I think the fact that most don’t read the paper from the tweet is fine. I don’t think anyone is changing an important, belief about something just from a tweet (and have even greater doubt that a scientist would). At best these tweets get filed back in the tweet reader’s brain under the “I think I vaguely remember that someone said X” folder. Whenever the tweet reader needs that knowledge, they presumably follow up by a more thorough review of the literature on the topic than the original tweet. The tweet is just there to flag that a study exists and claims an interesting result.

    PS: that regression to the mean idea is a really interesting result! But I didn’t read the paper yet, so I’ll just have to file that into my hypothetical “I think I vaguely remember that someone said X” folder 😉

  5. Somewhere in Population Biology of Plants (1977), John L. Harper supposedly stated that “all woodland herbs that have ever been studied are declining” — his point being exactly the same as yours, and that these populations usually begin after disturbances and peter out when the canopy closes.

Leave a Comment

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

WordPress.com Logo

You are commenting using your WordPress.com 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.