The “Anscombe’s quartet” is a famous way to illustrate how datasets can have nearly identical descriptive statistic but very different distributions. It serves as a warning, calling researchers to carefully visualize their data in any data analysis process.

More recently, Lucy D’Agostino McGowan has proposed the “Causal quartets”: a group of four datasets all with the same statistical summaries & visualizations but different true causal effects. See her blog post and paper preprint.

• Takeaway: theory-informed data analysis is key. A good theoretical understanding of the causal mechanisms behind a phenomenon is critical if we want to correctly draw causal paths & estimate true causal effects.