It's true for confidence intervals for difference of proportions, the samples should be independent. But the language feels off. How would men and women in each sample not be independent of each other?