Aggregation Math Example 1

Follow the full solution, then compare it with the other examples linked below.

Example 1

medium
Simpson's Paradox: Hospital A has a 90% recovery rate overall. Hospital B has 85%. However, for severe cases: A has 70%, B has 75%; for mild cases: A has 98%, B has 95%. Explain the paradox.

Solution

  1. 1
    Hospital B is better for BOTH severe (75%>70%) and mild (95% vs 98%... wait: A=98>95=B) — let's check: for mild cases A=98%, B=95%, so A is better for mild
  2. 2
    Revised: A better for mild (98% vs 95%), B better for severe (75% vs 70%)
  3. 3
    Overall: A=90% > B=85% — A wins overall despite B winning for severe cases
  4. 4
    Paradox: A's higher overall rate is because A sees mostly mild cases (high baseline rate); B sees more severe cases (dragging its average down); comparing without accounting for case mix is misleading

Answer

Simpson's Paradox: aggregated rates can reverse when a confounding variable (case severity mix) is ignored.
Simpson's Paradox occurs when aggregated data reverses the direction seen in subgroups. The overall rate is a weighted average where the weights (case mix) differ between groups. Aggregation can hide critical information about composition.

About Aggregation

Aggregation is the process of combining many individual data values into a single summary statistic such as a sum, mean, count, or proportion.

Learn more about Aggregation →

More Aggregation Examples