Empirical Likelihood Confidence Interval of a Difference in Upper Percentiles Between Two Groups

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Jihnhee Yu*+ and Albert Vexler and Alan D Hutson
• Companies: SUNY at Buffalo and SUNY at Buffalo and SUNY at Buffalo
• Keywords: Empirical Likelihood ; Quantile ; Confidence Interval

Abstract

In the group comparison of biomarkers, a difference in upper percentiles may be of specific interest as the upper percentile represents the upper range of biomarkers or used as the cut-off value for a disease classification. In this paper, we theoretically investigate confidence interval calculation for a quantile difference based on the empirical likelihood methodology. Two approaches, the classical empirical likelihood and 'plug-in' empirical likelihood are investigated. Extensive simulation results demonstrate the method based on the 'plug-in' estimation preforms slightly better in terms of coverage rate and the mean interval width, while the empirical likelihood approaches in general provide confidence interval estimations much more robust to various underlying distributions than methods based on normal approximation.