Another problem is that scientists usually publish only a small proportion of their data. If they cherry-pick the results that suit their hypotheses, this will introduce another source of bias, sometimes called ‘publication bias’, and sometimes called the ‘file-drawer effect’, because negative results are left in files
