My entire day consisted of making this figure:

I spent the day learning what principle component analysis is, how to do it, and what type of analysis my data needs. Now I get to figure out what this means. I will say that the principle component scores are correlated with my physiological function of interest. The black dots are babies, red are children, and green are adult individuals of my study species.

I had to do all of this at the behest of my primary collaborator. He's in town for a couple of months (he's been in Israel for two years for his postdoc) and the first thing he did was tell me that my multiple linear regressions are all wrong because I can't assume that my variables are independent. This is his way of getting around that, I guess. Instead of plugging my lipid groups into the MLR individually, I'll use the principle component scores. This, of course, means all new statistics and all new graphs.

I said to him, "You know I hate you right now, right?"

"It will pass," he replied.

My perspective on vaccines

9 hours ago in The Phytophactor

hey, you are already *on* a PC :-) (genetically)

ReplyDeleteI actually credit reading your blog with why I was able to pick up on the idea behind this kind of plot so quickly.

ReplyDeleteYou should include the eigenvalues for each dimension (assuming you didn't warp the graph to account for the eigenvalues).

ReplyDeleteFor the publication, yeah. This is just Minitab output, I'll wind up re-drawing it in SigmaPlot later.

ReplyDelete