Is there a bias when using effect sizes in favor of lower achieving students?