The problem with scientific studies is that they're great at describing a population, but generally have very little predictive power at the individual level.

For example, one of the studies that showed boys were better than girls at math showed that if you chose a random boy and a random girl, there was a 55% chance that the boy was better at math. Not very much different from flipping a coin at an individual level, but quite significant in a population.