Thanks Ralph, Moshe and [EMAIL PROTECTED] for you helpful comments.
Using bootstrap (e.g., 'boot' + boot.ci()) for the confidence interval on the variance is not very accurate in coverage, because the sampling distribution is extremely skewed. In fact, the 'BCa' method returns the same result as the Efron 'percent' method.
Moshe's idea of treating the confidence interval for the binomial variance as a transform of the confidence interval for the binomial proportion is elegant (Doh! Why didn't I think of that?), except that the transform is bivalued, although monotonic on each branch, with the branch point singularity at p=0.5.
The bootstrap method does not have much coverage accuracy for any proportion, for n=6, 12 and 20, and the proportion method works great for n=6, 12, 20 and 50, except near p = 0.5, where it fails to achieve reasonable coverage.
So I'm still looking for a reliable method for all p and for reasonable n. The proportion-based method is the best I've found, so far. ================================================================ Robert A. LaBudde, PhD, PAS, Dpl. ACAFS e-mail: [EMAIL PROTECTED] Least Cost Formulations, Ltd. URL: http://lcfltd.com/ 824 Timberlake Drive Tel: 757-467-0954 Virginia Beach, VA 23464-3239 Fax: 757-467-2947 "Vere scire est per causas scire" ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
