Simpson's Paradox

I recently observed something strange during the college admission process of one of my close friends. Before the admission process, it was known that the admission rate for men was higher than the admission rate for women. The institution concluded that this was because men were more likely to succeed in the admission process compared to women.

However, when the data was analyzed separately for different departments, it was found that women were more likely to get admitted than men. This seemed paradoxical, as the overall trend showed a higher admission rate for men.

After some discussion and research, I learned that a similar phenomenon had occurred in the Berkeley Gender Bias Controversy in the 1990s, which gave rise to the concept of Simpson's Paradox.

Simpson's Paradox is a phenomenon where the data trends observed in different subgroups can be reversed or obscured when the data is aggregated. This happens when there are lurking or confounding variables that are not taken into account.

In the case of my friend's college admission, the lurking variable might be the selection criteria used by individual departments. These department-level selection criteria could be affecting the admission rates for men and women in ways that are not evident when looking at the overall data.

The key takeaway is that when analyzing data, it's important to consider the possibility of lurking variables that can lead to counterintuitive results, as seen in the case of conflicting admission rates.