Is the use of effects sizes contingent on a normal distribution?