Contents

• 1 Introduction
  • 1.1 What is BUGS?
  • 1.2 For what kind of problems is BUGS best suited?
  • 1.3 Markov Chain Monte Carlo (MCMC) techniques
  • 1.4 A simple example
  • 1.5 Hardware platforms
  • 1.6 Software
  • 1.7 Format of manual
  • 1.8 Referring to BUGS
  • 1.9 Changes since Version 0.30
• 2 Getting started
  • 2.1 Obtaining BUGS and CODA
  • 2.2 Specific platforms
    • 2.2.1 Sparc Stations
    • 2.2.2 HP's
    • 2.2.3 PCs
  • 2.3 Using BUGS with S-plus
  • 2.4 Testing the installation
• 3 Specifying your model
  • 3.1 Directed graphical models
  • 3.2 Restrictions to log-concave distributions
• 4 Expressing your model in the BUGS language
  • 4.1 Lexical conventions
  • 4.2 Declarations
  • 4.3 Model description
  • 4.4 Distributions
  • 4.5 Using multivariate nodes
  • 4.6 Bounds on the domain of variables
  • 4.7 Functions
  • 4.8 Modelling functions of parameters
  • 4.9 Log-concave priors
  • 4.10 Indexing
    • 4.10.1 Functions as an index
    • 4.10.2 Implicit indexing
    • 4.10.3 Nested indexing
  • 4.11 for loops
  • 4.12 Functions of variables
  • 4.13 Data transformations within BUGS
• 5 Getting your data into BUGS: data and initial value files
  • 5.1 Data and initial value files
  • 5.2 Forward sampling to generate initial values
  • 5.3 File format for data and initial value files: rectangular format
  • 5.4 File format for data and initial value files: S-plus format
  • 5.5 Variable-length records (ragged arrays)
  • 5.6 Missing data
  • 5.7 Random number seed
  • 5.8 Preparation of data using statistical packages
• 6 Running BUGS
  • 6.1 BUGS commands
  • 6.2 Running BUGS in the background
  • 6.3 Example
  • 6.4 Errors when running BUGS
• 7 Processing output
  • 7.1 Main output
  • 7.2 Output of summary statistics
  • 7.3 Convergence: using the diag command
  • 7.4 Convergence: using the out command
• 8 Errors
  • 8.1 Errors found by BUGS
  • 8.2 Errors found by the operating system
  • 8.3 Undetected errors
• 9 Topics in modelling
  • 9.1 General Strategy
  • 9.2 `Non-informative priors'
    • 9.2.1 Issues with improper priors
    • 9.2.2 Pareto priors for precision parameters
    • 9.2.3 Assessment of proper priors for precision parameters
  • 9.3 Model criticism and selection
    • 9.3.1 Checking through examination of individual observations.
    • 9.3.2 Comparison between two or more candidate models
    • 9.3.3 Global goodness-of-fit tests based on Bayesian p-values
  • 9.4 Implementation of selected model checking criteria in BUGS
  • 9.5 Ranking
  • 9.6 Measurement error
  • 9.7 Multinomial-Poisson transformation
    • 9.7.1 Multinomial-logistic models
    • 9.7.2 Conditional likelihoods in case-control studies
    • 9.7.3 Partial likelihoods in Cox regression models
  • 9.8 Logical constraints
  • 9.9 Parameterisation
  • 9.10 Undirected links, chain graphs and neighbourhood models
  • 9.11 Multivariate normal nodes
    • 9.11.1 Use of Multivariate normal observations
    • 9.11.2 Use of Multivariate normal parameters
    • 9.11.3 Use of the Wishart distribution
    • 9.11.4 Avoiding multivariate nodes for parameters
• Acknowledgements
• References

Comments to bugs@mrc-bsu.cam.ac.uk
© 1995 MRC Biostatistics Unit
Return to the Welcome Page