APPROXIMATE FIDUCIAL INFERENCE FOR BINOMIAL DISTRIBUTIONS BASED ON THE LOGIT TRANSFORMATION
二项分布基于logit变换的近似信仰推断

In this paper, an approximate pivotal equation is constructed for a binomial distribution based on asymptotic normality of a modefied-logit-transformed estimator. Then an approximate fiducial distribution of the binomial parameter is obtained. Numerical results for one-sample cases and two-sample cases show that the interval estimators derived by the approximate fiducial distribution have better finite-sample frequentist properties than those suggested in the literature.