Abstract
This paper aims to investigate the performance and reliability of different regression methods in a value relevance model. Ten regression methods, both linear and nonlinear, were investigated to assess the effects of outliers, sample size, and overfitting on the model's performance (R² and error). The findings revealed that the Ordinary Least Squares (OLS) method is susceptible to false positive results due to its high R² but is often affected by heteroscedasticity, is more sensitive to outliers, sample size, and overfitting compared to the other regression methods analyzed. Although OLS is widely used in research in the field, it may not adequately address research questions in value relevance. The research advocates for a broader use of advanced regression techniques, including machine learning, to enhance empirical studies in finance. This research offers important insights for regulators, academia, companies, investors, and other stakeholders on the use of value relevance studies for decision-making.
© 2018, Facultad de Contaduría y Administración, Universidad Nacional Autónoma de México. All rights reserved. Publication of the article implies full assignment of property rights (copyright) in Journal of Contaduría y Administración. The publication mreserves the right to total or partial reproduction of the work in other print, electronic or any other alternative means, but always recognizing its responsibility.
License for Published Content
Unless otherwise stated, all contents of the electronic edition of the journal are distributed under a license and distribution "Creative Commons Attribution-Noncommercial 4.0 International" (CC-by). You can see from here the version of the license information. This circumstance must be expressly stated in this way when necessary.
Metadata License
The metadata of papers published by Contaduría y Administración are in the public domain, through the publisher's waiver of all rights to the work under copyright law worldwide, including all rights and related rights, to the extent permitted by law. You may copy, modify, and distribute the metadata, even for commercial purposes, without requesting permission.
