The appropriate threshold for claiming association depends also on the context and the relative costs for false positive and false negative results. For example, re-sequencing a region and conducting in vivo and in vitro functional studies is quite expensive, and will require convincing evidence that the observed association is true. On the other hand, including a region in a predictive genetic risk score is relatively inexpensive, so a less stringent threshold might suffice. This approach to replication is intuitively Bayesian (although it need not use formal Bayesian methods): each successive study serves to update the prior for association in subsequent studies.
