Selecting on statistical significance and practical importance is wrong [pdf]

stat.columbia.edu