"...warned against drawing conclusions based on simple correlation."

Can we just do this, in general?