Simpson's Paradox

Thanks to Rob Campbell, one of the Harvard students who is working with me as a teaching assistant (“mentor”) in the Crimson Summer Academy, for pointing me to this wonderful example of Simpson’s Paradox: a statistics page at SUNY Oswego demonstrates a scatterplot with a surprisingly negative correlation between high-school SAT scores and college GPAs!

When the scores are disaggregated into the three distinct colleges whose results had been lumped together, we end up with three strong positive correlations instead of one negative one. It’s worth thinking about why the negative correlation appears when the scores are combined.

Categories: Math