Fig 3 Funnel plot of effect size by inverse standard error, showing no evidence of publication bias in studies of ethnicity and academic performance in trained doctors and medical students