A longest run test for heteroscedasticity in univariate regression model

TL;DR

This paper introduces a simple, general test for detecting heteroscedasticity in univariate regression models that does not assume normality or regularity, showing favorable simulation results.

Contribution

It presents a novel, easy-to-compute heteroscedasticity test that is more general than existing nonparametric methods.

Findings

Test performs well compared to other nonparametric tests

No assumptions on error normality or response function regularity

Applicable to a wide range of univariate regression models

Abstract

The scope of this paper is the presentation of a test that enables to detect heteroscedasticity in univariate regression model. The test is simple to compute and very general since no hypothesis is made on the regularity of the response function or on the normality of errors. Simulations show that our test fairs well with respect to other less general nonparametric tests.