You are viewing a single comment's thread from:
RE: How small should p-values be in the life sciences?
Nice. What about publishing all the results, not just the 10 where the null hypothesis was (correctly or not) rejected? If we publish all 50 experiments, then about 4-5 of them would be erroneous due to statistical chance, which is the 5% error rate we were willing to accept.
Publishing all of the studies would not help with the false positive rate itself, but it would help immensely in making accurate decisions about what we should set the p threshold at. If we want a certain false positive rate in a field, we can engineer that rate if we know going in what is the probability that an alternative hypothesis is true. That probability is really hard to determine when all studies are not published, and this is a separate and contributing issue to replication crises in certain fields.