The consequences of selective participation on behavioral-genetic findings: evidence from simulated and real data