Diesel Acceleration
ø
Spent the morning at Diesel and made good headway on the plot of disparity measures. I think the three main measures (mean pairwise distance, convex hull volume, and alpha shape volume) are mostly in place now. Now, I just need to check other measures, Doug Erwin’s paper in Palaeontology a couple of years back comes to mind, perhaps throw together a quick plot of variance or range on each axis (?), and—importantly—add a panel for diversity. Oh, and combine all the panels together on one graphics device/PDF.
I got a little worried about taxon sampling when I was making the disparity plots this morning—I had been operating under the “range-through” model in the function I wrote to return taxon lists for each time bin, and I got curious as to what it would like like under “in-bin” sampling. The results were, as you might expect, quite a bit different (though I should run a formal comparison on the completed figure that has the assembled panels). This was some brief cause of concern to me but I realized that the key to doing this right is to compare measures of diversity with morphospace constructed with the same taxon lists. What’s more interesting (I hope) is how diversity and morphospace are related given the same taxon sampling, rather than comparing, say, SQ-subsampled diversity to range-through morphospace. Ideally, diversity—morphospace comparisons under the different taxon sampling models will yield similar results in the sense that the relationship between the two will be the same, even if the outcomes compared amongst the models are quite different. Otherwise, it would be important (and interesting) to comment on how taxon sampling affects the story. Complicated, confusing, but I think that’s the right approach.
In any case, here are the panels from this morning:
A crucial (because very vexing!) detail on the bottom plot is the greek character alpha, something I’ve been trying to do over and over again over the past few months and only just figured out. Very rewarding. It turns out there’s a {plotmath} package (apparently part of the base installation) with a function bquote() that does the trick, beautifully. The only downside is that, unlike most things in base R, it’s not vectorized—so you can’t pass a vector variable to it and have it plot all of the values in it, so you have to use a for loop. Ugly, but not a problem.
In the course of the afternoon, consolidated the plots into a four-panel figure and started working on the fourth panel, diversity. Didn’t quite finish, but got close, and felt very happy with the amount of work I got done and the level of productivity and focus. Once I’d gotten home and cooked dinner (it was a rather elaborate effort), though, I was exhausted, so the workday did not stretch into the evening this time. But I am still satisfied. Hitting the cafés is definitely a good idea, and I really think I need to do it more often.
