This paper presents four methods of constructing the confidence interval for the proportion p of the binomial distribution. Evidence in the literature indicates the standard Wald confidence interval for the binomial proportion is inaccurate, especially for extreme values of p . Even for moderately large sample sizes, the coverage probabilities of the Wald confidence interval prove to be erratic for extreme values of p . Three alternative confidence intervals, namely, Wilson confidence interval, Clopper-Pearson interval, and likelihood interval , are compared to the Wald confidence interval on the basis of coverage probability and expected length by means of simulation.
展开▼