CURRENT RESEARCH TOPICS

More Powerful Unconditional Inference on Treatment Effect from Binary Matched Pairs 2. We consider the problem of testing for a difference in the probability of success from matched binary pairs. We point out that the probability of a discordant pair is limited by the process of matching provided only that the treatment does not reverse the association of the response and the matching variable. This extra mild assumption means that maximization with respect to the nuisance parameter does not have to range over the entire intervals [0,1]. This results in more efficient P-values which are never larger and often smaller than the more general P-values. Exact lower/upper limits are also larger/smaller than limits obtained using full maximization. An early draft of this paper is HERE.