Contents

• Introduction
  • What is CODA?
  • A cautionary note on convergence diagnostics
  • Routines implemented in CODA version 0.30
  • Hardware/Software requirements
  • Referring to CODA
• Getting started
  • Obtaining CODA
  • Installation
    • Within UNIX
    • Within DOS
  • Testing the installation
    • Generating BUGS output for testing CODA
    • Using CODA to analyse the line output
• Output Analysis
  • Plots
    • Printing graphical output from CODA
    • Multiple pages of plots
    • Displaying multiple graphics windows on screen
  • Statistics
• Convergence Diagnostics
  • Geweke
    • Plotting Geweke's diagnostic
  • Gelman & Rubin
    • Plotting Gelman & Rubin's diagnostic
  • Raftery & Lewis
  • Heidelberger and Welch
  • Autocorrelations
    • Autocorrelation plots
  • Cross-correlations
    • Cross Correlation plots
• Changing the CODA defaults
  • Selecting variables for analysis
  • Selecting chains for analysis
  • Selecting iterations for analysis
  • Selecting the thinning interval
  • Default settings for plots
    • Selecting which plots to produce
    • How to plot each chain on a separate graph
    • Controlling the page layout of CODA plots
    • Specifying the bandwidth function for the kernel density estimator
    • Producing a postscript file of the graphical output from CODA
    • Specifying the exponent size required for producing plot labels in scientific notation
  • Default settings for summary statistics
    • Combining chains to produce a single summary for each variable
    • Setting the batch size
    • Selecting the posterior quantiles to be estimated
    • Controlling the number of significant digits displayed for text output
  • Default settings for convergence diagnostic tests
    • Changing the window size for Geweke's diagnostic
    • Changing the bin width for Geweke's and Gelman & Rubin's diagnostics
    • Specifying the parameters for Raftery & Lewis' diagnostic
    • Specifying the precision for Heidelberger and Welch's halfwidth test
    • Combining chains for cross-correlation analysis
• Exiting from CODA and saving the output
• Using CODA with Gibbs sampling output from other programs
• References
• Index
• About this document ...