Matched-pair data are employed in various clinical study designs. Agresti and Min [1] indicated that a Wald-type statistic confidence interval difference of marginal probability shows that the actual confidence level becomes lower than the nominal confidence level and proposed a confidence interval by adding k times the hypothetical trials to the Wald interval. In this study, we propose a new confidence interval based on Agresti’s method concerning the ratio of marginal probability. Furthermore, we indicate the usability of the confidence interval, which we suggest in this paper through a simulation.

matched-pair data, confidence interval, cross-over design, binary data.