On confidence limits for the difference of two binomial parameters

Bayesian sample‐size determination for inference on two binomial populations with no gold standard classifier

“…If the odds ratio is of interest rather than the di erence , then the computational burden of the HPD interval may be worth the e ort because the odds ratio posterior requires a larger sample size to be symmetric and unimodal. The ALC [6,21] for a (1 − )100 per cent interval of width l requires a sample size n such that n n12=0 n n11=0 l (n 11 ; n 12 )m(n 11 ; n 12 )6l (2) where l (n 11 ; n 12 ) denotes, in our case, the length of an equal-tailed (1 − )100 per cent posterior interval of the di erence . That is, (2) is a weighted average of the lengths of all possible 95 per cent credible intervals where the weights are marginal probabilities of the data.…”

“…The comparison of two proportion parameters is one of the most commonly used procedures in statistical inference. Much work continues to be done on the problem, including recent papers by Pan [1], Feigin and Lumelskii [2], Railkar et al [3], Agresti and Ca o [4], and Cai [5]. The design of experiments to yield inferences on two proportions has also been discussed in the literature quite frequently.…”