A Graphical Diagnostic for Identifying Influential Model Choices in Bayesian Hierarchical Models

.  Real-world phenomena are frequently modelled by Bayesian hierarchical models. The building-blocks in such models are the distribution of each variable conditional on parent and/or neighbour variables in the graph. The specifications of centre and spread of these conditional distributions may be well motivated, whereas the tail specifications are often left to convenience. However, the posterior distribution of a parameter may depend strongly on such arbitrary tail specifications. This is not easily detected in complex models. In this article, we propose a graphical diagnostic, the Local critique plot, which detects such influential statistical modelling choices at the node level. It identifies the properties of the information coming from the parents and neighbours (the local prior) and from the children and co-parents (the lifted likelihood) that are influential on the posterior distribution, and examines local conflict between these distinct information sources. The Local critique plot can be derived for all parameters in a chain graph model.

[1]  Bradley P. Carlin,et al.  Bayesian measures of model complexity and fit , 2002 .

[2]  D. J. Spiegelhalter,et al.  Identifying outliers in Bayesian hierarchical models: a simulation-based approach , 2007 .

[3]  D. Gaver,et al.  Robust empirical bayes analyses of event rates , 1987 .

[4]  Michael I. Jordan.  Graphical Models , 2003 .

[5]  M. J. Bayarri,et al.  P Values for Composite Null Models , 2000 .

[6]  R. Wolpert,et al.  Combining models of health and exposure data: the SAVIAH study , 2001 .

[7]  D. Rubin.  Bayesianly Justifiable and Relevant Frequency Calculations for the Applied Statistician , 1984 .

[8]  Bent Natvig,et al.  A Robust Conflict Measure of Inconsistencies in Bayesian Hierarchical Models , 2007 .

[9]  A. F. Smith,et al.  Conjugate likelihood distributions , 1993 .

[10]  Geir Storvik,et al.  Posterior Predictive p‐values in Bayesian Hierarchical Models , 2009 .

[11]  D J Spiegelhalter,et al.  Approximate cross‐validatory predictive checks in disease mapping models , 2003, Statistics in medicine.

[12]  I. Guttman.  The Use of the Concept of a Future Observation in Goodness‐Of‐Fit Problems , 1967 .

[13]  LA MORTALITE PAR CANCER EN FRANCE , 1997 .

[14]  V. Johnson.  Bayesian Model Assessment Using Pivotal Quantities , 2007 .

[15]  Valen E. Johnson,et al.  Comment: Bayesian Checking of the Second Levels of Hierarchical Models , 2007, 0802.0749.

[16]  Xiao-Li Meng,et al.  POSTERIOR PREDICTIVE ASSESSMENT OF MODEL FITNESS VIA REALIZED DISCREPANCIES , 1996 .

[17]  Adrian F. M. Smith,et al.  Sampling-Based Approaches to Calculating Marginal Densities , 1990 .

[18]  Applications of the Local critique plot , 2010 .

[19]  Xiao-Li Meng,et al.  Posterior Predictive $p$-Values , 1994 .

[20]  N. Hjort,et al.  Post-Processing Posterior Predictive p Values , 2006 .

[21]  D. Ankerst,et al.  Kendall's Advanced Theory of Statistics, Vol. 2B: Bayesian Inference , 2005 .

[22]  P. Green,et al.  Hidden Markov Models and Disease Mapping , 2002 .

[23]  J. Besag,et al.  Bayesian image restoration, with two applications in spatial statistics , 1991 .

[24]  George E. P. Box,et al.  Sampling and Bayes' inference in scientific modelling and robustness , 1980 .

[25]  Dipak K. Dey,et al.  A simulation-intensive approach for checking hierarchical models , 1998 .

[26]  R. Wolpert,et al.  Integrated likelihood methods for eliminating nuisance parameters , 1999 .

[27]  Andrew Thomas,et al.  WinBUGS - A Bayesian modelling framework: Concepts, structure, and extensibility , 2000, Stat. Comput..

[28]  Michael Evans,et al.  Checking for prior-data conflict , 2006 .

[29]  H. Moshonov.  Checking for Prior-Data Conßict with Hierarchically SpeciÞed Priors , 2005 .

[30]  M. Kendall,et al.  Kendall's advanced theory of statistics , 1995 .

[31]  Bent Natvig,et al.  Extensions of a Conflict Measure of Inconsistencies in Bayesian Hierarchical Models , 2009 .