Say, for instance, you believe redheads cause more traffic accidents than those with other colors of hair.
You can then divide the world into redheads and non-redheads and compare their accident records.
So you make sure that if your population of redheads had a million people with a certain distribution of age, the distribution in your non-redhead sample is exactly the same.
The computer reveals that redheads go to the ER more often and break bones more often.
And of the redheads themselves, are there factors among the clumsy ones that are different than the coordinated ones?
Is it actually that blue-eyed redheads have the same number of accidents as non-redheads, but brown-eyed redheads are even more clumsy, accident prone, and traffic hazards?