The final rows of Table 1 present the meta-analytic effect size averaged across all runs having the same sign of the first published study, i.e., negative or positive. One may wonder whether the selective publishing scenario yields a biased estimate of the effect depending on the sign of the first study. The results of our simulation show that meta-analyses with as few as about 8 published studies already yield an unbiased estimate, confirming W&H (p.3) that possible bias in selective publishing (as modeled by W&H) is rapidly nullified.
