If some percentage of panicky people have good coping mechanisms, but averaged over all panicky people, having a good coping mechanism is still less likely than

If some percentage of panicky people have good coping mechanisms, but averaged over all panicky people, having a good coping mechanism is still less likely than for non-panicky people, then it's still efficient to filter out all the panicky people. Probabilities matter and it's better if you pick from a group that is more likely rather than one which is less likely, even if you miss some good people in the less likely group.