|Path diagram to figuring out what type of multivariate statistics to use.|
Dave Wedin use to joke that the Joe Craine way of analyzing the data is to take it all and put it into a PCA and see what you get.
There's some truth to that. PCA has served me well. And almost every time I get forced to use a different multivariate analysis, PCA seems to give me similar results.
For example, in the last burning paper a reviewer insisted we use NMDS for our plant cover data rather than PCA. Correlation coefficient between NMDS and PCA Axes 1-3: 0.93, 0.89, 0.71. Same story from each.
Still, it's good to know what other options there are out there.
I've been looking at bacterial data with Noah and found the Ramette 2007 paper on multivariate analyses in microbial ecology. Noah uses Principal Coordinates Analysis and I was trying to remember the difference between PCA and PCoA.
It's a good user guide to multivariate statistics in general.
One thing that was interesting was a multivariate analysis of the different types of multivariate statistics different disciplines use.
Ramette, A. 2007. Multivariate analyses in microbial ecology. FEMS Microbiology Ecology 62:142-160.