Re: Kieren, thank you for avoiding profanity this time
@ Doctor Syntax: OK perhaps for those of a more simple disposition I should have spelt it out.
Science in general but specifically the science of genetics.
Sorry to correct you, group traits are one thing, however they are totally useless for finding the specific profile(s) you might need to perform specific roles. So for insurance companies assessing the likelihood that you as a new customers might have a bad accident profile, statistics are imperfect yet valuable. However, for someone seeking a brain surgeon, random group statistics would lack any semblance of value the statistics would need hugely more care.
In one country where I worked, the female recruits were infinitely better programmers than their male counterparts, based not upon group based guesses, or doubtful statistics, but upon both training course results (>90% female pass rates vs <35% male pass rates) and work output. Since this was in a country with no history of female employment, (we broke the model) the statistics were simply not available - beyond previously confirming that most previous male recruits did not meet the requirements.
It is vanishing rare to need a person for a role and simply obtain them from a group assessment as the above example shows.