Bayesian modeling of variance heterogeneity in normal regression models. (English) Zbl 0983.62013

Cepeda, Edilberto; Gamerman, Dani

Braz. J. Probab. Stat. 14, No. 2, 207-221 (2000).

Summary: This paper considers the situation where regression models are proposed for the mean and the variance of normal observations. We initially summarize the classical approach for modeling variance heterogeneity in normal regression analysis. Next, we provide an exposition of the MCMC algorithm proposed to draw approximate samples from the resulting posterior distribution. We illustrate this algorithm with simulated data and apply it to the minitab tree data, comparing it with the classical analysis of this dataset. The paper is concluded with a few proposed extensions.

MSC:

62F15	Bayesian inference
62J05	Linear regression; mixed models

Keywords:

link functions; working observations; Markov chain Monte Carlo; Metropolis-Hastings algorithm

Software:

GLIM; bnormnlr