Question

A hospital analyst compares two treatment programs. Looking at the combined dataset, Program A has a higher overall recovery rate than Program B. However, when the analyst splits the data by patient severity (mild vs. severe cases), Program B actually has a higher recovery rate within BOTH the mild group and the severe group. What phenomenon is the analyst observing, and what should guide the conclusion?

Accepted Answer

Simpson's paradox; the segmented (subgroup) results are more valid because a lurking variable—case severity—is confounding the aggregated comparison

Answer

Sampling error; the analyst should increase the sample size until the aggregate and subgroup results agree

Answer

Multicollinearity; the analyst should drop the severity variable to resolve the contradiction

Answer

Regression to the mean; the aggregated result is correct and the subgroup differences will disappear over time