Space-varying regression models : specifications and simulation

Dani Gamerman, Ajax R.B. Moreira, Håvard Rue

Space-varying regression models are generalizations of standard linear models where the regression coefficients are allowed to change in space. The spatial structure is specified by a multivariate extension of pairwise difference priors thus enabling incorporation of neighboring structures and easy sampling schemes. Different sampling schemes are available and may be used in an MCMC algorithm. These schemes are compared in terms of chain autocorrelation and resulting inference. We also discuss different prior specifications that accommodate the spatial structure. Results are illustrated with simulated data and applied to a real dataset.

MoreLess

Year of publication:	January 2015
Authors:	Gamerman, Dani ; Moreira, Ajax ; Rue, Håvard
Publisher:	Brasília, DF, Brazil : Ipea, Institute for Applied Economic Research
Subject:	Bayesian \| Hyperparameters \| Gibbs sampling \| Markov chain Monte Carlo \| Markov random fields \| Metropolis-Hastings algorithm \| Sampling schemes \| Markov-Kette \| Markov chain \| Theorie \| Theory \| Stichprobenerhebung \| Sampling \| Monte-Carlo-Simulation \| Monte Carlo simulation \| Simulation \| Bayes-Statistik \| Bayesian inference \| Regressionsanalyse \| Regression analysis