Proper multivariate conditional autoregressive models for spatial data analysis

Abstract

In the past decade conditional autoregressive modelling specifications have found considerable application for the analysis of spatial data. Nearly all of this work is done in the univariate case and employs an improper specification. Our contribution here is to move to multivariate conditional autoregressive models and to provide rich, flexible classes which yield proper distributions. Our approach is to introduce spatial autoregression parameters. We first clarify what classes can be developed from the family of Mardia (1988) and contrast with recent work of Kim et al. (2000). We then present a novel parametric linear transformation which provides an extension with attractive interpretation. We propose to employ these models as specifications for second-stage spatial effects in hierarchical models. Two applications are discussed; one for the two-dimensional case modelling spatial patterns of child growth, the other for a four-dimensional situation modelling spatial variation in HLA-B allele frequencies. In each case, full Bayesian inference is carried out using Markov chain Monte Carlo simulation.

Is Poverty Decentralizing? Quantifying Uncertainty in the Decentralization of Urban Poverty

“…This is a multivariate CAR (MCAR) prior, and is similar to that proposed by Gelfand and Vounatsou (2003). Such models are typically specified in their conditional form; that is, as a series of distributions for f k conditional on the remaining random effects.…”

“…Here r is a spatial dependence parameter, with r close to one corresponding to strong spatial dependence in the data and r D 0 corresponding to independence in space. We note that this formulation implies the same level of spatial dependence (same value of r) for each census year, but this simplification could be relaxed if needed (for details, see Gelfand and Vounatsou [2003] and the wider MCAR literature). Relaxing this assumption, however, increases model complexity unnecessarily, and we justify it for our data in the next section.…”

Multivariate Parametric Spatiotemporal Models for County Level Breast Cancer Survival Data

“…As in the univariate case, if jaj < 1, B is positive definite (as long as K is), but we again typically take 0 < a < 1. Besides being a propriety parameter, a can be interpreted as a coefficient which measures spatial association for each spatial random process (Gelfand and Vounatsou, 2003). If a ¼ 0, the / i are independent across the strata.…”

“…That is, it does not appear sensible to incorporate spatial association into the q i , b 0i , and g i mixing distributions, but ignore the correlation among these three components in any given stratum. Fortunately the multivariate conditionally autoregressive (MCAR) distribution, introduced by Mardia (1988) and recently shown to be computationally feasible for hierarchical modeling by Gelfand and Vounatsou (2003) and , emerges as ideal for rectifying this problem. In Section 2 we review CAR distributions in both the univariate and multivariate cases.…”

Default Bayesian analysis for multivariate generalized CAR models

“…Multiple response variables are available as indicators of health status, and as a result, models for multivariate spatial lattice data are an indispensable tool for analyzing health disparity data. Recently, Greco and Trivisano (2009), Zhang et al (2009), Jin et al (2007), Sain and Cressie (2007), Jin et al (2005), Gelfand and Vounatsou (2003), Carlin and Banerjee (2003) and Kim et al (2001) explored multivariate spatial models for lattice data, adopting the Bayesian framework as the natural inferential approach. The only exception, Sain (2009) developed the maximum likelihood estimation procedure for a special case, the multivariate gaussian conditional autoregressive (CAR) model of Sain and Cressie (2007).…”