An Improved Method for Constructing Confidence Intervals for Quantiles from Survey Data

• Activity Number: 639
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Akhil Vaish*+ and Babubhai Shah
• Companies: RTI International and SAFAL Institute
• Keywords: Francisco and Fuller ; Quantile estimation ; Confidence intervals ; Complex survey data

Abstract

The Francisco-Fuller (1991) approach yields unsatisfactory results when applied to data from multistage unequal probability samples. There have been problems encountered in practical application of the Francisco-Fuller method: it requires evaluation of bounds at many points, the resulting limits are, at times, not monotonic, and the stated coverage is not very accurate at the boundaries. An adjustment to the empirical distribution is used to achieve internal consistency and reduce coverage error. Hypothetical populations will be generated using normal as well as some skewed distributions and 1000 simulation samples will be drawn from each population. Multiple confidence intervals with varying confidence levels will be generated for several percentiles for each of the sample by two methods. The coverage rates obtained from the proposed method and the Francisco-Fuller will be compared.