TLDR: Correlation != causation; if you have high dimensional data, you can always find a correlation, but it's probably meaningless; polynomial wiggle is a bitch, so don't fit high dimensional polynomials to your data.