I’m going to talk about population biology today, more of that below.
But first, being new to the weblog lark, I was genuinely chuffed when those poor misguided fools lovely folks at the NN coalface office were kind enough to give me a forum to vent my spleen. I promised I would use the opportunity to talk about so much more than being a Brit abroad. Apparently us expats are using up all the oxygen ether online space with our constant whinging and griping about how hard it is to get a good cuppa.
So, before I spit in the eye of the NN lovelies, by telling them I’m off on holiday and will probably be slacking off on the posting front for the next couple of weeks while I enjoy a few ampolles de Moritz in Barcelona, my first visit to Mallorca (where I’ll pop in to see these charming folks) and perhaps some delicious dairy products in Chizé, I thought I’d start off my proper science weblogging career with a “Spot the bloody obvious mistake” competition.
Spot the bloody obvious mistake – a competition
A recent article, by Eberhardt et al, published in a reputable Scandanavian journal of ecology, aims to establish which of 5 pretty simple population growth models (related to the Logistic growth model) best describes natural population time series. So far, so seems to be a reasonable thing to do.
The data comes from an interesting resource, the Global Population Dynamics Database (GPDD) and there have been a few papers trawling through this recently to try and make generalizations about population growth rates and shapes of density dependence etc. I’ll go into some of the issues with this sort of approach in more detail when I’m back from my travels (spatial structure, species interactions, anyone?), but here’s something to get your brainboxes ticking over:
In fig. 1 of the paper, some data is plotted from a Beaver population, allegedly from Missouri (time series 177 in the GPDD), but unfortunately the real (GPDD) data doesn’t look anything like what’s presented in the paper. For this reason, I’ll show 2 plots below, the first is the GPDD data, the 2nd is my loose approximation1 of the data presented in the paper (not sure about copyright issues, so I won’t copy the original image).
A travesty, earlier this time period
Neither the referees, nor handling Editor (who is a good guy to go for a beer with) seem to have checked this or asked why the data presented in fig. 1 doesn’t look anything like the data in the database, The other beaver data presented in this paper (GPDD time series 172) doesn’t match the data from the GPDD either. I stopped checking after that, so I’m not sure whether the deer, fox, or elephant seal data matches. The figure caption in the paper contains little useful information (little = 0) about any data transformations that may have been done, but I’ve no idea why they would transform the data anyway, and the y-axis labels say something like “Population numbers”. (sorry, can’t access the pdf just now to check).
Assuming I’m not making an embarrassing mistake here (go on, please, just for once?), why did this happen? Should referees (or editors) be expected to check freely available data sources when these data-trawling exercises can deal with thousands of time-series? This paper uses less than 15 I think. Should I write to the journal and moan at them about this? I’m a postdoc, and still need all the publications I can get! Perhaps it would be more sensible to contact the author first to find out what’s going on?
Apart from these issues, there’s something particularly fishy (pun may or may not be intended) about the Eberhardt et al beaver population data.
Answers on an electronic postcard, to the address below. First prize will be a glowing sense of pride in your excellent work.
1 It’s not exactly the same as the journal fig. data, but close enough for the point to be clear. I honestly have no idea where their plotted data came from, and I don’t have my data-grabber software with me to grab their data.
On copyright, you should be OK to put up the graph, as you’re using it to comment on: it counts as fair use.
As a referee, I wouldn’t check the data unless something seemed fishy anyway (i.e. other aspects of the paper suggested the authors didn’t know what they were doing. Or did know rather too well). I think we have to assume honesty unless there’s a good reason not to.
I would comment on the graph (it does look a bit too neat), but the pit stops that are coming up are more interesting.
I’m one of the people moaning about cups of tea (though I am a Brit living in the UK, where you can’t get a decent cup…..oh there I go again, sorry). So I probably shouldn’t comment further.
However, I broadly agree with Bob’s take. Nature journals are possibly different from the one you describe here, in that N journals have professional editors (who aren’t practicing academics) so we do ask our referees to look at all the data carefully. However, in this era of mega, “terror” and peta, we also understand the time and effort constraints for the referees, also. (Certainly, any data that underlie the submission should be available to be checked.) We do ask referees to let us know if they think there may be something not quite right, as the onus is on the authors to convince rather than the refs to go through every tiny detail.
I don’t know how to answer your question – if it were a Nature journal I would know, I’d suggest you first contact the author and then if no joy submit a technical comment. The journal editors would then look into it, in other words, ask the authors to respond in a timely fashion, then see what the referees thought of the exchange (sometimes, under these circumstances, authors cave in and we go straight to a published corretion; sometimes, the person complaining turns out not to be justified; other times it is both interesting and a grey area, so we publish the exchange, linked to the original publication).
I do agree with Bob that the system is set up on the basis that people are honest. But at the same time, there are good referees and not so good referees, one reason why we use two or three (sometimes more) per submission, depending on the expertise required to judge it.
At the end of the day, the published record should be as correct as we can all make it, so I do hope you do inform the author and journal. It might be time-consuming, but if there is a mistake here, the authors will be grateful to you for pointing it out so they can make the correction.
Thanks for the feedback. Contacting authors certainly seems to me to be the polite thing to do, but there is still the nagging temptation to write something directly to the journal. I may put up the original figs once I’ve recovered from the jet lag…
I looked at the figures, and they all seem a bit too neat. If you’re still in the Land of the Sour Coffee on Wednesday, we could have a chat about it. I can also pass on all those salacious unsubstantiated rumours about Nature editors that I
madepicked up whilst back home.Sounds just like the kind of
gossipuseful info I´d love to hear about. Unfortunately, I´ve already started enjoying the Barcelona beers and will be in Mallorca on wednesday. Let´s chat when I return in a couple of weeks though.I can also confirm that the y-axis labels in the original MS were, simply, “Counts”.
Give Daniel my regards. I’ve been intending to email him about some data, but I’ve been strategically waiting until other stuff gets done.