Course contentsShow
← Data Science
Lesson 1428 of 2,14536. Causal Inference BasicsFree lesson

The Simpson's Paradox Example

Understand how aggregated data can show reversed relationships when confounders are ignored.

The Simpson's Paradox Example

You'll learn how aggregated data can completely reverse the direction of a relationship when confounding variables lurk beneath the surface.

What is Simpson's Paradox?

Simpson's Paradox occurs when a trend that appears in several groups of data disappears or reverses when those groups are combined. It's one of the most striking demonstrations of why confounding variables matter in causal inference.

A Classic Example: University Admissions

Imagine a university where overall admission rates show:

  • Men: 45% admitted
  • Women: 30% admitted

This suggests gender bias, right? But look at department-level data:

Department A (competitive):

  • Men applicants: 200, admitted: 20%
  • Women applicants: 100, admitted: 30%

Department B (less competitive):

  • Men applicants: 100, admitted: 80%
  • Women applicants: 300, admitted: 70%

Women had higher admission rates in both departments! The paradox emerges because women applied disproportionately to the harder department (A), while men applied more to the easier one (B). Department choice is the confounder.

Why This Matters

When you aggregate data without accounting for confounders, you can draw completely wrong conclusions. The overall numbers suggested discrimination against women, but disaggregated data revealed the opposite pattern within each department.

This isn't just academic—Simpson's Paradox appears in medical studies, business metrics, and A/B tests whenever subgroups have different baseline characteristics.

Key Takeaway: Always disaggregate your data by potential confounding variables before drawing causal conclusions—what's true in aggregate may reverse at the group level, and vice versa.