Siliceous and Delicious

Although the rain caused a bit of a delay this morning (it was so disgusting that, even wearing full waterproofs, walking in turned out to be so unpleasant as to be not worth the effort—but the bus, of course, zipped by, full), I have been blazing forward. I finished up the character/PCO association plot, which involved hand-crafting a legend (the legend() function does not work with the symbol() function I used to plot the bubbles), realizing in the process that the color scale was off, that I needed to generate my own palette, and thus that I needed to learn the basics of R color palettes and how they’re used. It’s all fairly low-level stuff.

In any case, I was mighty excited when I had a final product in hand, and skipped into Andy’s office to show him—he was also very excited and said “this should definitely be a figure in your paper”, so I suppose it was worth the extra two days it took to make it. Here’s what the final thing looks like.

This plot lets me say two things:  (a) most of the strongest and most significant associations between the original characters and the PCO axes are in the first few axes, so the 2-D plots will be somewhat meaningful (i.e. big and dark circles are toward the left), and (b) there is also a fair amount of information contained in the higher PCO axes, meaning that there is real complexity structure to the data beyond the first few axes, too, and some of the characters won’t show variation expressed clearly in the first few axes. This is good, because it means I wasn’t wasting my time adding all those other characters into the morphospace—they aren’t just covarying perfectly with other characters, but have a variance structure of their own.

With the Cramér analysis that the plot is generated from, I can also now extract a) which the most important variables are that contribute to axes 1, 2, and 3, and b) which the most important variables are on other axes, that won’t show up on those plots. This is what I need to do next.

For PCO1, the ten characters with the largest Cramér values are:

X60 X102 X89 X2 X61 X90 X91 X26 X41 X1

0.81 0.51 0.43 0.43 0.40 0.39 0.38 0.35 0.35 0.34

For PCO2, they are:

X34 X26 X60 X51 X102 X84 X12 X63 X2 X27

0.81 0.49 0.40 0.38 0.36 0.32 0.31 0.30 0.29 0.28

And for PCO3:

X102 X68 X70 X73 X34 X91 X72 X60 X41 X12

0.41 0.38 0.36 0.35 0.35 0.34 0.34 0.32 0.29 0.27

Now, I’ll need to somehow highlight these on a PCO1-2 or PCO1-2-3 plot, to show how they fall on the plot, and also illustrate what they are (or at least a few of them). Altogether the biggest few are X60, X34, X102, X26, X89, X2, X61, and X60. Those are all above 0.4.

In the midst of my head spinning with all of these details and things to do, I decided I needed to step back and sketch out my plan for the main figures for the paper. I did this…

…and in the middle of it all, I received an email from the Freshman Dean’s Office informing me that I have made the first cut and am being asked for an interview! This was a fabulously exciting piece of news and came as a crowning moment for an already successful week. Time to run off for a well-deserved dinner. Huzzah!