Siliceous and Delicious

Another spectacularly late start to the day. Whatever it is that’s making me this tired (the heat? visitor fatigue?), I hope I’m able to shake it this weekend. Although we do have more visitors coming. Ah well.

At the end of the day, it’s productivity that counts, not hours, so time to shake the guilt over being late and time to get down to it. The next tractable task on my Gmail to-do list is plotting up the ‘three-timer’ preservation statistic for diatoms. I had completed the code for this the last time I worked on it (whenever that was—curiously, I don’t seem to have written about it on this blog), but couldn’t get it to work. So, debugging time it is.

I got the code to work reasonably well—I think!—although the result was not at all what I expected. It looks like preservation, as reported by this measure of the occurrence data itself, is fairly constant over time. This is quite surprising, because there is a discrepancy between the sampled in-bin and the range-through diversity curves, and the discrepancy (a similar, though not identical, quantity to the three-timer statistic) varies over time.

Here, for comparison, the preservation as recorded qualitatively in the Neptune database (as good [2], moderate [1] or poor [0]):

What exactly to make of this, I don’t know. Is the objective preservation data bogus? Have I made a mistake in calculating the three-timer statistic? Here’s the SIB/RT ratio, for comparison, which should be similar to the three-timer statistic, with the only exception (that I can think of) that SIB and RT also count 1- and 2-timers. But somehow they ought to be qualitatively similar. The fact that this graph is all over the place makes me think I might have done something wrong with the three-timer function:

Of course, the first-order shape of this graph (above) is a U, which is what’s expected even if preservation is uniform (but imperfect) through time. Because there’s more time either side of the middle, you’re more likely to identify missing taxa there. At the end points of the time series, you don’t know what taxa came before (at the beginning) or after (at the end), and so—by definition—your RT count is the same as your SIB count.