Title: Regression analysis using the imprecise Bayesian normal model

Authors: Lev V. Utkin

Addresses: Department of Computer Science, St. Petersburg State Forest Technical Academy, Institutsky per. 5, St. Petersburg, 194021, Russia

Abstract: A class of regression models taking into account the lack of sufficient statistical data is proposed. The main ideas of the class are to use the framework of Vapnik|s learning theory and to replace a single probability distribution of the noise by a set of probability distributions, which is not the parametric set of distributions. The set of probability distributions is defined by the imprecise Bayesian normal model. A numerical example illustrates the proposed regression models.

Keywords: imprecise probabilities; lower probability distributions; upper probability distributions; Vapnik; learning theory; risk; Bayesian inference; regression modelling; imprecise modelling; Bayesian normal model.

DOI: 10.1504/IJDATS.2010.037477

International Journal of Data Analysis Techniques and Strategies, 2010 Vol.2 No.4, pp.356 - 372

Published online: 14 Dec 2010 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article