Originally Posted by JonClarke

I saw a classic example of this at a conference two weeks ago. A research group did multi-variant correlation analysis in relationship to acid groundwater in a particular region. The best correlation was with vegetation, ergo the acidity was caused by the vegetation. They did not consider the opposite possibility, that the acidity was controlling the vegetation.

I have seen this happen time an again when people mine the data without understanding relationships, and think that greater computing power = better science.