Modified ridge and other regularization criteria: A brief review on meaningful regression models

Article type: Research Article

Authors: Lipovetsky, Stan

Affiliations: Minneapolis, MN, USA | E-mail: [email protected]

Correspondence: [*] Corresponding author: Minneapolis, MN, USA. E-mail: [email protected].

Abstract: The work describes a series of techniques designed to obtain regression models resistant to multicollinearity and having some other features needed for meaningful results. These models include enhanced ridge-regressions with several regularization parameters, regressions by data segments and by levels of the dependent variable, latent class models, unitary response, models, orthogonal and equidistant regressions, minimization in Lp-metric, and other criteria and models. All the approaches have been practically implemented in various projects and found useful for decision making in economics, management, marketing research, and other fields requiring data modeling and analysis.

Keywords: Ridge-regressions, regularizations, latent class regression, equidistant model, Lp-metric

DOI: 10.3233/MAS-210536

Journal: Model Assisted Statistics and Applications, vol. 16, no. 3, pp. 225-227, 2021