Continuity Corrected Wilson Interval for the Difference of Two Independent Proportions

Abstract Confidence interval for the difference of two proportions has been studied decades. Many methods were developed to improve approximation limiting distribution test statistics, such as profile likelihood method, score and Wilson method. For by Beal (Biometrics 43:941, 1987), Z statistic standard normal may be further improved utilizing continuity correction, in observation anti-conservative intervals from interval. We theoretically prove that is nested corrected under mild conditions. compare with commonly used regards coverage probability, width, mean squared error probability. The proposed good performance many configurations. An example a Phase II cancer trial illustrate application these methods.