Donner A
J Chronic Dis. 1982;35(6):429-35. doi: 10.1016/0021-9681(82)90057-1.
Problems of interpretation often arise when patients who have been entered into a randomized clinical trial are post-stratified according to prognostic descriptors, and treatment comparisons performed within subgroups. In this paper a Bayesian approach to interpreting these comparisons is proposed, which takes into account the investigator's prior expectations concerning the relative magnitudes of treatment effects in different subgroups.
