Review Articles

Graph-based multivariate conditional autoregressive models

Ye Liang

Department of Statistics, Oklahoma State University, Stillwater, Oklahoma, USA

ye.liang@okstate.edu

Pages 158-169 | Received 25 Nov. 2018, Accepted 07 Sep. 2019, Published online: 16 Sep. 2019,
  • Abstract
  • Full Article
  • References
  • Citations

ABSTRACT

The conditional autoregressive model is a routinely used statistical model for areal data that arise from, for instances, epidemiological, socio-economic or ecological studies. Various multivariate conditional autoregressive models have also been extensively studied in the literature and it has been shown that extending from the univariate case to the multivariate case is not trivial. The difficulties lie in many aspects, including validity, interpretability, flexibility and computational feasibility of the model. In this paper, we approach the multivariate modelling from an element-based perspective instead of the traditional vector-based perspective. We focus on the joint adjacency structure of elements and discuss graphical structures for both the spatial and non-spatial domains. We assume that the graph for the spatial domain is generally known and fixed while the graph for the non-spatial domain can be unknown and random. We propose a very general specification for the multivariate conditional modelling and then focus on three special cases, which are linked to well-known models in the literature. Bayesian inference for parameter learning and graph learning is provided for the focused cases, and finally, an example with public health data is illustrated.

References

  1. Atay-Kayis, A., & Massam, H. (2005). A Monte Carlo method for computing the marginal likelihood in nondecomposable Gaussian graphical models. Biometrika92, 317–335. doi: 10.1093/biomet/92.2.317 [Crossref][Web of Science ®], [Google Scholar]
  2. Banerjee, S., Gelfand, A., & Carlin, B. (2004). Hierarchical modeling and analysis for spatial data. Boca Raton, FL: CRC Press/Chapman & Hall. [Google Scholar]
  3. Besag, J. (1974). Spatial interaction and the statistical analysis of lattice systems. Journal of the Royal Statistical Society. Series B36, 192–236. [Google Scholar]
  4. Dobra, A., Lenkoski, A., & Rodriguez, A. (2011). Bayesian inference for general Gaussian graphical models with applications to multivariate lattice data. Journal of the American Statistical Association106, 1418–1433. doi: 10.1198/jasa.2011.tm10465 [Taylor & Francis Online][Web of Science ®], [Google Scholar]
  5. Gelfand, A., Sahu, S., & Carlin, B. (1995). Efficient parameterizations for normal linear mixed models. Biometrika82, 479–488. doi: 10.1093/biomet/82.3.479 [Crossref][Web of Science ®], [Google Scholar]
  6. Gelfand, A., & Vounatsou, P. (2003). Proper multivariate conditional autoregressive models for spatial data analysis. Biostatistics4, 11–25. doi: 10.1093/biostatistics/4.1.11 [Crossref][Web of Science ®], [Google Scholar]
  7. Green, P. (1995). Reversible jump Markov chain Monte Carlo computation and Bayesian model determination. Biometrika82, 711–732. doi: 10.1093/biomet/82.4.711 [Crossref][Web of Science ®], [Google Scholar]
  8. Jin, X., Banerjee, S., & Carlin, B. (2007). Order-free co-regionalized areal data models with application to multiple-disease mapping. Journal of the Royal Statistical Society: Series B (Statistical Methodology)69, 817–838. doi: 10.1111/j.1467-9868.2007.00612.x [Crossref][Web of Science ®], [Google Scholar]
  9. Jin, X., Carlin, B., & Banerjee, S. (2005). Generalized hierarchical multivariate CAR models for areal data. Biometrics61, 950–961. doi: 10.1111/j.1541-0420.2005.00359.x [Crossref][Web of Science ®], [Google Scholar]
  10. Kim, H., Sun, D., & Tsutakawa, R. (2001). A bivariate Bayes method for improving estimates of mortality rates with a twofold conditional autoregressive model. Journal of the American Statistical Association96, 1506–1521. doi: 10.1198/016214501753382408 [Taylor & Francis Online][Web of Science ®], [Google Scholar]
  11. Letac, G., & Massam, H. (2007). Wishart distributions for decomposable graphs. The Annals of Statistics35, 1278–1323. doi: 10.1214/009053606000001235 [Crossref][Web of Science ®], [Google Scholar]
  12. MacNab, Y. (2011). On Gaussian Markov random fields and Bayesian disease mapping. Statistical Methods in Medical Research20, 49–68. doi: 10.1177/0962280210371561 [Crossref][Web of Science ®], [Google Scholar]
  13. MacNab, Y. C. (2016). Linear models of coregionalization for multivariate lattice data: A general framework for coregionalized multivariate CAR models. Statistics in Medicine35, 3827–3850. doi: 10.1002/sim.6955 [Crossref][Web of Science ®], [Google Scholar]
  14. MacNab, Y. C. (2018). Some recent work on multivariate Gaussian Markov random fields. Test27, 497–541. doi: 10.1007/s11749-018-0605-3 [Crossref][Web of Science ®], [Google Scholar]
  15. Mardia, K. (1988). Multidimentional multivariate Gaussian Markov random fields with application to image processing. Journal of Multivariate Analysis24, 265–284. doi: 10.1016/0047-259X(88)90040-1 [Crossref][Web of Science ®], [Google Scholar]
  16. Martinez-Beneito, M. (2013). A general modeling framework for multivariate disease mapping. Biometrika100, 539–553. doi: 10.1093/biomet/ast023 [Crossref][Web of Science ®], [Google Scholar]
  17. Martinez-Beneito, M. A., Botella-Rocamora, P., & Banerjee, S. (2017). Towards a multidimensional approach to Bayesian disease mapping. Bayesian Analysis12, 239–259. doi: 10.1214/16-BA995 [Crossref][Web of Science ®], [Google Scholar]
  18. Scott, J., & Berger, J. (2006). An exploratory of aspects of Bayesian multiple testing. Journal of Statistical Planning and Inference136, 2144–2162. doi: 10.1016/j.jspi.2005.08.031 [Crossref][Web of Science ®], [Google Scholar]
  19. Scott, J., & Carvalho, C. (2009). Feature-inclusion stochastic search for Gaussian graphical models. Journal of Computational and Graphical Statistics17, 790–808. doi: 10.1198/106186008X382683 [Taylor & Francis Online][Web of Science ®], [Google Scholar]
  20. Wang, H., & Li, S. (2012). Efficient Gaussian graphical model determination under G-Wishart prior distributions. Electronic Journal of Statistics6, 168–198. doi: 10.1214/12-EJS669 [Crossref][Web of Science ®], [Google Scholar]
  21. Wang, H., & West, M. (2009). Bayesian analysis of matrix normal graphical models. Biometrika96, 821–834. doi: 10.1093/biomet/asp049 [Crossref][Web of Science ®], [Google Scholar]