CURRENT RESEARCH TOPICS

Exact confidence limits for treatment effect in a 2x2 table with structural zero. For a correlated 2x2 table where the (01) cell is empty by design, the parameter of interest is typically the ratio of probability of secondary success conditional on primary success to the probability of primary success. We look at constructing exact upper limits using the method of Buehler (1957) and using as generating statistics five different basic templates. The results are favourable to exact limits based on the SRLR statistics. There are stark asymmetries in the results for upper limits and lower limits. This is joint work with max Moldovan and an early draft of this paper is HERE. A shorter version is under revision at Biometrical Journal