Photo of Guiyuan Lei

Dr Guiyuan Lei
University of Newcastle
Newcastle upon Tyne, United Kingdom

Calibayes CISBAN

Home  Research  Resource  iNotes  Personal

Resource

Biological networks

Monte Carlo

Bioinformatics

Misc

Graphical Gaussian Model:

• Gaussian Papers

  1. "Sparse graphical models for exploring gene expression data", Adrian Dobra, Chris Hans, Beatrix Jones, J.R.Joseph R. Nevins, Guang Yao and Mike West, Journal of Multivariate Analysis 90 (2004) 196-212.

  2. "Decomposable graphical Gaussian model determination", Paolo Giudici; Peter J. Green, Biometrika Vol. 86, No. 4 (Dec., 1999), pp. 785-801
  3. "Hyper Markov Laws in the Statistical analysis of decomposable graphical models", A.P. Dawid and S.L. Lauritzen, The Annals of Statistics 1993, Vol. 21, No.3, 1272-1317
  4. "EXPERIMENTS IN STOCHASTIC COMPUTATION FOR HIGH-DIMENSIONAL GRAPHICAL MODELS" (manuscript), Beatrix Jones, Carlos Carvalho, Adrian Dobra, Chris Hans, Chris Carter & Mike West, Statistical Science 20 (2005): 388-400

  5. "A Monte Carlo method for computing the marginal likelihood in nondecomposable Gaussian graphical models", Aliye Atay-Kayis and Hélène Massam, Biometrika 2005 92: 317-335
  6. "Bayesian inference for nondecomposable graphical Gaussian models", Petros Dellaportas, Paolo Giudici and Gareth Roberts, Sankhya: The Indian Journal of Statistics 2003, Volume 65, Pt. 1, 43--55
  7. "Hyper inverse wishart distribution for non-decomposable graphs and its application to Bayesian inference for gaussian graphical models", Alberto Roverato, Scandinavian Journal of Statistics Vol. 29: 391-411, 2002

  8. "Parameter Priors for Directed Acyclic Graphical Models and the Characterization of Several Probability Distributions", Dan Geiger and David Heckerman, The Annals of Statistics 2002, Vol. 30, No.5, 1412-1440
  9. "Learning Gaussian Networks", Dan Geiger and David Heckerman, Microsoft Research Technical Report (MSR-TR-94-10), July 1994, p28
  10. "An empirical Bayes approach to inferring large-scale gene association networks", Juliane Schäfer and Korbinian Strimmer, Bioinformatics, vol 21, pages 754-764, 2005
  11. "Comparative evaluation of reverse engineering gene regulatory networks with relevance networks, graphical Gaussian models and Bayesian networks", Adriano V. Werhli, Marco Grzegorczyk, and Dirk Husmeier, Bioinformatics, July 14, 2006

    12. For time course data:

  12. Learning causal networks from systems biology time course data: an effective model selection procedure for the vector autoregressive process. BMC Bioinformatics 8 Suppl. 2: S3, Opgen-Rhein, R., and K. Strimmer. 2007. (Preprint). This paper uses an analytic shrinkage approach to estimate VAR regression coefficients.
  13. "Process Pathway Inference via Time Series Analysis", C. H. Wiggins and I. Nemenman, Experimental Mechanics, Vol 43, No. 3, 361-370. This paper constructs appropriate priors for Bayesian inference of process pathway based on exploiting well-established knowledge about transcriptional regulation. This goal, constraining possible models with biologically motivated prior is useful because the genomic time series data are usually scarce. The model used here is AR(1) model.
  14. "Bayesian Analysis of econometric time series models using hybrid integration rules", Dani Gamerman and Ajax R. B. Moreira, Comunications in Statistics: Theory and Methods, Vol 31, No. 1, 2002. 49-72. This paper is concerned with the study of Bayesian inference procedures to commmon components multivariate dynamic linear model (DLM) including: univariate dynamic linear model, time-varying VAR models and structural VAR models. Inference procedures are based on a hybrid integration scheme where state parameters are analytically integrated and hyperpaprameters are integrated by Markov chain Monte Carlo methods. Normal inverse Wishart priors are used here.
  15. "Dynamic Matrix-Variate Graphical Models", Carlos M. Carvalho and Mike West, To appear in Bayesian Analysis. This paper introduces a novel class of Bayesian models for multivariate-time series analysis based on synthesis of dynamic linear models and graphical models. The systhesis uses sparse graph modelling ideas to introduce structured, conditional independence relationships in the time-varying, cross-sectional covariance matrices of multiple time series. A new class of models and their theoretical structure involving novel matrix-normal/hyper-inverse Wishart distributions is defined. The resulting Bayesian methodology and computational strategies for model fitting and prediction is then described.
  16. "Priors and Component Structures in Autoregressive Time Series Models", Gabriel Huerta and Mike West. This paper focuses on decomposition and root structure for autoregressive time series models, introduces new approaches to prior specification and sructuring.
  17. "Searching for the causal structure of a vector autoregression", Demiralp S and Hoover K. D., Oxford Bull. Econom. Statist. 2003, 65:745-767. Vector autoregressions (VARs) are economically interpretable only when identified by being transformed into a structural form (the SVAR) in which the contemporaneous variables stand in a well-defined causal order (See page 2 of this paper for the definition of causal order). This paper applies graph-theoretic methods of searching for causal structure based on relations of conditional independence to select among the possible causal orders. The paper uses the PC algorithm (book "Causation, Prediction, and Search" by Spirtes, Peter, Clark Glymour, and Richard Scheines. Published by Cambridge, MA: MIT Press) embeded in Spirtes et al.'s TETRAD 3 software. The paper contributes two things. First, provides an accessible account of the underlying rationale for the graph-theoretic approach to causal order in general. Second, using a simulation study, they address the efficacy of the most common algorithm for implementing this approach to selecting the causal order of SVARs.
  18. "Graphical models for structural vector autoregressions", Moneta A, 2004. Technical Report. Laboratory of Economics and Management, Sant Anna School of Advanced Studies, Pisa. This paper modified the PC algorithm which was developed by Spirtes, Glymour and Scheines. Original PC considers computional complexity issues, it is possible to recover sparse graphs with as many as a hundred variables (Spirets et al. "Causation, Prediction, and Search", 2000, p.87). This paper focuses on stability, in the sense small errors of the algorithm input (conditionally independence tests) are likely to produce less large errors of the algorithm output (casual relationships), with respect to the original PC algorithm. VAR models of macroeconomic time series deals with a very limited number of variables.
  19. "Bayesian estimates for vector autoregressive models", Ni S., Sun D., J. Business Economic Statist. 2005, 23:105-117. This paper examines frequentist risks of Bayesian estimates of VAR regression coefficient and error covariance matrices under competing loss functions, under a variety of non-informative priors, and in the normal and Student-t models.

• Useful Links

Bayesian Networks:

• Tutorial

• Software

 

Last modified:
11 October, 2007