Quantile information is useful in business and engineering applications, but the exact sampling distribution of sample quantile is often unknown. In this paper, we study the performance of four nonparametric methods, the kernel density estimation (KDE), bootstrap percentile (BP), bootstrap-t (BT) and accelerated bias-correction bootstrap (BCa) methods, through Monte Carlo simulations for conducting interval inference on the quantiles of normal and generalized Pareto distributions. Simulation results show that the BCa and BP methods outperform the BT and KDE methods. Sample sizes to implement the recommended nonparametric methods for inferring a range of upper quantiles are also studied based on the coverage probability.
Inference on the distribution quantiles using nonparametric.... Available from: https://www.researchgate.net/publication/319095266_Inference_on_the_distribution_quantiles_using_nonparametric_methods [accessed May 09 2018].