Document Type

Pre-Print

Publication Date

6-26-2025

Subjects

Regression analysis -- Mathematical models, Multivariate regression -- Analysis

Abstract

Portnoy (2022) introduced a version of Canonical Correlation anal- ysis based on Regression Quantiles. The proof of the asymptotic convergence of the estimators contained some gaps. A new proof is developed using Bahadur representations to obtain asymptotic represen- tations for the Canonical Regression Quantile components with error uniformly Op (n-1/4 log n). These innovative representations justify approximations and automatically provide joint convergence in dis- tribution. This enables an inductive argument to obtain asymptotic results for all components. These novel Bahadur approximations may have more general application and interest.

Rights

Copyright 2025 Stephen Portnoy

This is an author manuscript made available under a CC-BY 4.0 International license.

Persistent Identifier

https://archives.pdx.edu/ds/psu/44002

Download

Share

COinS