[sampling algorithms0]    Sampling Algorithms
   
Version 2.10, April 2005
      
   Andrew Thomas
1
      

   
1 Dept of Mathematics & Statistics,
      University of Helsinki
      Helsinki
      Finland


   e-mail: ant@rni.helsinki.fi   
   
Contents

    Choice of Sampling Algorithms
   
    Description of Sampling Algorithms

Choice of Sampling Algorithm [top]

When OpenBUGS starts up a file called
methods.txt in /Updater/Rsrc is read. The modules implementing the sampling methods listed in this file are then loaded dynamically. Factory objects to create updater objects which use these methods are stored in an array (in the same order as in the file). The first factory object in this array is used to create updaters for all the stochastic nodes in the model that need updating and for which the associated sampling method is valid and wishes to create updaters. If there are any nodes which do not have updater objects created by this first factory the second factory is tried and so on. In this way only a single updater object is created for each node that needs sampling. If non of the factories are able to create an updater object for a node in the model that needs sampling an error is reported. It is the function of the factory to decide if its associate sampling method is valid for a particular node in the model. Factory objects associated with sampling algorithms listed at the start of the methods file can impose strict conditions on the node for which they are able to create an updater object. In general algorithms at the start of the methods file do block updating of nodes. Users can develope special purpose sampling algorithms and place them at the start of the methods file. Algorithms towards the end of the method file tend to be general purpose (and somewhat less efficient) than earlier algorithms.

The methods file can contain comments. If the algorithm occurs after a comment symbol (the #) it is not loaded or used. An algorithm might be commented out because it is less efficient than a competing algorithm, because it performs badly in some test situations, because it causes less efficient algorithms to be chosen for other nodes or because it has not been tested enough. Users are enouraged to try some of the commented out algorithms and to report their good and bad experiences.

The module UpdaterKrigparam is given as an example of special purpose updater that will only be used for sampling the parameters of a kriging prior. The actual sampling algorithm used is the general purpose delayed single component adaptive metropolis but the factory object only returns an updater object if the likelihood is a kriging prior.

Description of Sampling Algorithms [top]

Here is an alphabetical list of install procedures for updaters that can be used followed by a brief note on how the algorithm works and what type of situation it is appropiate to.
   
UpdaterAMblock.InstallGLM

   Current point Metropolis type updater with delayed rejection and continously adapted multivariate normal proposal distribution. For a fixed effect block of nodes all of which have a a logistic or log-linear conditional distribution. Non markov algorithm. Not currently used.

UpdaterAMblock.InstallHetro

   Current point Metropolis type updater with delayed rejection and continously adapted multivariate normal proposal distribution. For a fixed effect block of nodes the first element of which has a generic conditional distribution. Non markov algorithm. Not currently used.
   
UpdaterAMblock.InstallHomo

   Current point Metropolis type updater with delayed rejection and continously adapted multivariate normal proposal distribution. For a fixed effect block of nodes all of which have a generic conditional distribution. Non markov algorithm. Not currently used.
   
UpdaterAMblock.InstallRE

   Current point Metropolis type updater with delayed rejection and continously adapted multivariate normal proposal distribution. For a random effect block of nodes all of which have a logit or log-linear conditional distribution. Non markov algorithm. Not currently used.
   
UpdaterAMNLLS.InstallMarginal   
   Continously adaptive block metropolis algorithm for non-linear least squares problem with the measurement error precision integrated out.
   
UpdaterAMNLLS.InstallCond   
   Continously adaptive block metropolis algorithm for non-linear least squares problem.
   
UpdaterBeta.Install

   Gibbs type of updater for beta conditional distribution.
   
UpdaterCatagorical.Install    

   Gibbs type of updater for univariate node which takes descrete values with a upper bound. Works by enumeration. Slow if many catagories.
   
UpdaterChain.Install
   Updater for a node with chain graph prior. Works as a sequence of univariate updaters but can inpose a constraint such as the sum to zero constariant for CAR priors.
   
UpdaterDescreteSlice.Install   

   Slice type sampler for descrete variable. Competes with UpdaterCatagorical if descrete varaible has upper bound and with UpdaterMetbinomial if no upper bound.
   
UpdaterDFreeHybrid.Install

   Metropolis type of updater for a block of nodes which has a generic conditional distribution with unbounded support. Proposal distribution is based on integration of Hamiltons equations of classical mechanics. Derivatives of log likelihood are calculated numericaly. Not currently used. There are three broad classes of updater: those that do Gibbs sampling, those that do Metropolis Hastings sampling and those that do slice sampling.
   
UpdaterDirichlet.Install

   Gibbs type of updater for a conditional distribution that is dirichlet.
   
UpdaterDirichletprior.Install

   Slice sampler type of updater for a conditional distribution that has a dirichlet prior but non conjugate likelihood.
   
UpdaterForward.Install   

   Gibbs type of updater for univariate node which does not have any likelihood.
   
UpdaterGamma.Install

   Gibbs type of updater for gamma conditional distribution.
   
UpdaterGLM.InstallGLM

   Metropolis Hastings type updater for a block of nodes which have either a log-linear or logistic conditional distribution with unbounded support.

UpdaterGLM.InstallNormal

   Gibbs type of updater for multivariate conditional distribution where the prior is a set of univariate normal nodes. Not currently used.
   
UpdaterGriddy.Install    

Independene Metropolis Hastings type updater for generic univariate distribution.   Distribution must have bounded support. Builds a trapizoidal approximation to the conditional to use as the proposal. Slow algorithm. Not currently used, slice sampling is prefered but see note below about multimodality.

UpdaterHybridglm.Install
   Updater for random effect block of nodes with normal prior and loit, log-linear.

UpdaterHybridglm.InstallChain
   Updater for chain graph prior and loit, log-linear or normal likelihood.
       
UpdaterHybridglm.Install

      Metropolis type of updater for a block of nodes which has a normal, log-linear or logistic conditional distribution with unbounded support. Proposal distribution is based on integration of Hamiltons equations of classical mechanics. Used for random effects. Not currently used.
   
UpdaterMetbinomial.Install

   Current point Metropolis type updater for univariate node which takes descrete values with no upper bound.
   
UpdaterMetnormal.InstallDelayed

   Current point Metropolis type updater for generic univariate distribution with unbounded support. Uses a normal proposal distribution that is adapted during a tunning phase. Uses delayed rejection if first proposal is rejected.
   
UpdaterMetnormal.InstallStd

   Current point Metropolis type updater for generic univariate distribution with unbounded support. Uses a normal proposal distribution that is adapted during a tunning phase.
   
UpdaterMultinomial.Install   

   Independence Metropolis utype of updater for a conditional distribution that has a multinomial prior.
   
UpdaterMVNLinear.Install

   As for UpdaterMVNormal.Install but the likelihood can have terms that are multivariate normal of a different dimension to the prior and the link function does not have to be the identity but can have a linear form.

UpdaterMVNormal.Install

   Gibbs type of updater for multivariate normal conditional distribution where each term in the likelihood is either normal, log normal or multivariate normal of same dimension as the prior with a unit link function.
   
UpdaterNormal.Install

   Gibbs type of updater for univariate normal conditional distribution.
   
UpdaterPareto.Install    

   Gibbs type of updater for pareto conditional distribution.
   
UpdaterPoisson.Install

   Gibbs type of updater for poisson prior with single binomial likelihood term having order equal to the prior.

UpdaterRejection.Install

   Gibbs type of updater for univariate node with log-linear or logistic distribution and unbounded support.
   
UpdaterSCAM.InstallDelayed

   Current point Metropolis type updater for generic univariate distribution with unbounded support. Uses a continously adapted normal proposal distribution. Non markov algorithm.   Not currently used. Uses delayed rejection if first proposal rejected.
   
UpdaterSCAM.InstallDelayedB

   Current point Metropolis type updater for generic univariate distribution with bounded support. Uses a continously adapted normal proposal distribution. Non markov algorithm. Not currently used. Uses delayed rejection if first proposal rejected.
   
UpdaterSCAM.InstallStd

   Current point Metropolis type updater for generic univariate distribution with unbounded support. Uses a continously adapted normal proposal distribution. Non markov algorithm. Not currently used.

UpdaterSlice.Install    

Slice type of updater for generic univariate distribution. Uses steping out search procedure to find the slice. Step length of search procedure adapted during tunning phase. Can miss modes for multimodal conditional distribution if the support is not bounded.

UpdaterWishart.Install   

   Gibbs type of updater for conditional distributions that is a wishart.