A practical procedure to obtain confidence intervals for the Bernoulli parameter (Q2266313)

The authors recommend four methods for obtaining a confidence interval (CI) for the parameter of the Bernoulli distribution considering optimality, average length, actual confidence level and computational simplicity [see the last two authors, A basic statistical problem: Confidence interval for the Bernoulli parameter. ibid. to appear]. In large sample situations all these four methods of choice may be represented from the practical point of view by just one, the Anscombe CI. The user can recognize such large sample situations from a table. Practical formulas for the Anscombe's CI are given [cf. \textit{F. J. Anscombe}, Biometrika 35, 246-254 (1948; Zbl 0032.03702)]. On the other hand for small sample situations the paper contains tables with the Sterne- Crow CI's for the 95\% and 99\% confidence levels obtained by using an algorithm given by \textit{E. L. Crow}, ibid. 43, 423-435 (1956; Zbl 0074.140).