Publication bias could have inflated our replicability estimates [29], [30]. Due to the large number of SNPs that are tested, the usual practice when publishing a GWAS has been to report all newly discovered associations, plus the replication status of previously associated SNPs. However, this is not always the case and, in some cases, not all previous results are discussed in each publication. Therefore, our ability to gather replication attempts depends on how many of them are explicitly reported, which presents enormous variability among papers. This opens the possibility of a reporting bias, in which GWAS authors could prioritize mentioning successful replication attempts, while overlooking failed replications. If so, our chance of gathering a replication attempt might be heavily biased towards positive results, thus inflating our estimates of replicability [30].
