You can see this is a problem if you mine out the distribution of p-values from articles. Andrew Gelman had a great post on this topic I can't find now. Pre-r

You can see this is a problem if you mine out the distribution of p-values from articles. Andrew Gelman had a great post on this topic I can't find now. Pre-registration could be a great solution. Negative results are also important.