This doesn't exclude overfitting. You can take samples n much larger than parameters p in, for example, a political or household earnings survey performed in th

This doesn't exclude overfitting. You can take samples n much larger than parameters p in, for example, a political or household earnings survey performed in the San Francisco Bay Area, but obviously one cannot safely assume no overfitting occurs in that scenario.