Is an effect size or p-value more important in replication studies?