Skip to content

A Generative High Quantile Homogeneity Test Using Bahadur Representation for Heteroskedastic High Quantile Regression of Tail Dependent Time Series

Home / Papers / A Generative High Quantile Homogeneity Test Using Bahadur Representation for Heteroskedastic High Quantile Regression of Tail Dependent Time Series

A Generative High Quantile Homogeneity Test Using Bahadur Representation for Heteroskedastic High Quantile Regression of Tail Dependent Time Series

Authors: Ting Zhang, Fangwei Wu, Jingying Gao Date: 2026-05-11 Paper ID: openalex:2605.10915

Summary

This paper introduces a formal framework for testing the homogeneity of covariate effects across different high quantiles in tail-dependent time series. The authors derive a novel Bahadur representation, which enables the reduction of nonlinear high quantile regression estimators to linear forms with controlled error bounds. This representation facilitates the construction of a generative homogeneity test capable of handling both homoscedastic and heteroskedastic conditions. The method is validated through comprehensive experiments on both synthetic and real-world datasets.

Key Contributions

  • Develops a novel Bahadur representation for high quantile regression estimators under heteroskedasticity and tail dependence.
  • Proposes a generative high quantile homogeneity test that assesses whether explanatory variables have uniform effects across tail quantiles.
  • Establishes theoretical foundations for converting nonlinear high quantile regression inference into linear form analysis with explicit error bounds.

Open Questions & Future Work

Key Concepts

Archivist Review

The approved concepts and open questions address fundamental theoretical bottlenecks in extreme value statistics and quantile regression. I have limited approvals to the central theoretical contribution (Bahadur representation) and the primary unresolved inferential limitation noted in the paper, ensuring these entries remain high-level and broadly relevant for future statistical time series research.

Approved Concepts

  • Bahadur representation for high quantile regression: It provides the theoretical foundation for reducing complex nonlinear high quantile regression estimator problems into manageable linear forms.

Approved Open Questions

  • Unified Inference in High Quantile Regression: This is a fundamental bottleneck in high quantile statistics where asymptotic validity often depends on specific, often unverifiable, assumptions about the rate of extremeness, hindering the reliability of automated or general-purpose statistical software for tail analysis.

Metadata & Links

url
https://arxiv.org/abs/2605.10915
paper_id
2605.10915
paper_source
openalex
domain
time-series
tags
time-seriesstatistical-inferencequantile-regression
architectures
datasets
concept_slugs
bahadur-representation-high-quantile-regression
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-05-14T06:09:54Z