Many thanks Ben, this is really helpful. Phyll > On 26 Jul 2014, at 13:48, "Djulbegovic, Benjamin" <[log in to unmask]> wrote: > > Phyll, > There is currently no consensus how large the effect size should be to overcome concerns of bias ( and random error). Paul Glasziou has suggested 10 times differences, GRADE suggests multiple studies with effects size 2-5 times as large ( going in the same directions). > We have been trying to answer this question empirically, but so far this has been hampered by lack of funding... > Hope this helps > Best > Ben > > Sent from my iPad > ( please excuse typos & brevity) > >> On Jul 26, 2014, at 8:41 AM, "Phyll Buchanan" <[log in to unmask]> wrote: >> >> Any thoughts on how to decide whether the effect sizes are so large they should overcome these concerns?