I need at least a couple of keywords to get me started and read for my self.
Suppose I have a number $M$ of populations. From each population, I draw $R$ replicas of a parameter vector $\vec x$ of $N$ dimensions. In total, I have $R\times M$ samples of $N$-dimensions each.
I want to see what components of $\vec{x}$ are different across across the populations. I would like a way to detect the most significant variations. Is this possible?