669 – Variance Estimation in Sample Surveys
Confidence Intervals for Quantile Estimation from Complex Survey Data
Babubhai Shah
SAFAL Institute Inc.
Akhil K. Vaish
RTI International
The estimation of confidence intervals for quantile estimation involves three steps: (1) point estimate of cumulative distribution function (CDF), (2) Confidence interval for the estimated CDF, which is a binomial proportion, (3) a method for converting CDF intervals to quantile intervals. The point estimate for CDF in a simple random sample is well known i/(n+1). We provide an equivalent estimate using weights for survey data. There are several options for confidence intervals of binomial proportions and two approaches for converting intervals of CDF to those for quantiles: The estimating equations Woodruff [1952], or Complete estimation of confidence interval for the entire CDF Francisco and Fuller [1991]. These result in several combinations of methods that are very similar and not easy to discriminate.