Anscombe's Quartet consists of four datasets with nearly identical R² values but very different data structures. What does this demonstrate?