In 1973 a large American university looked at its own admission figures and found something that looked indefensible. Across the whole institution, men were admitted at a noticeably higher rate than women.
Then somebody broke the same data down by department, and the picture reversed. In most departments women were admitted at the same rate as men or a slightly higher one. Nothing had been miscounted. Both figures came from the same table.
The explanation is entirely arithmetic, and a small invented example shows it. Suppose one department admits sixty percent of applicants and another admits twenty. Eighty men apply to the easy department and twenty to the hard one; for women it is the other way round. Every applicant of either sex faces exactly the same chance in whichever department they chose — and yet fifty-two of the hundred men are admitted and twenty-eight of the hundred women.
Nobody was treated differently. The two groups had simply applied to different places, and the overall rate was an average of departments weighted by who applied where.
This is the whole of what a table question tests. A numerator means nothing until you know its denominator, and the same count can be divided by three different totals to give three contradictory answers, every one of them correct.