Tests and Confidence Intervals for the Difference and Ratio of Two Probabilities
两个概率之差与比值的检验和置信区间
Fisher's (1925) test can deal with the hypothesis of equal probabilities of success of two experiments, A and B. The test requires a fixed number of experiments of both kinds. If the experimental design is charged into randomizing the numbers of experiments A and B, other kinds of hypotheses can be tested exactly and exact confidence intervals can be constructed.