English   -  Français  

Résumé de communication

Résumé 263 :

Estimation of high conditional quantiles for heavy-tailed distributions
He, Xuming
University of Michigan

Quantifying how high or low quantiles change with covariates is of interest in a wide range of applications, and is intrinsically challenging due to data sparsity in the tails. Quantile regression, which is a convenient and natural approach to assessing the impact of a covariate at different quantiles of a response variable, can suffer from high variability at tails, especially for heavy-tailed distributions. In this talk, we explore new estimation methods for high conditional quantiles by first estimating the intermediate conditional quantiles in a conventional quantile regression framework, and then extrapolating these estimates to the high tails based on reasonable assumptions on tail behaviors. We use asymptotic theory as well as empirical studies to demonstrate the potential of the proposed approach. When applied to statistical downscaling of daily precipitation in the Chicago area, this approach provides more stable results than the conventional quantile regression in the projection of extreme or nearly extreme precipitation in the area. The talk is based on recent work with Huixia Wang (North Carolina State University, USA) and Deyuan Li (Fudan University, China).