createMixtureTarget: Mixture target distribution

Description Usage Arguments Value Author(s) References See Also

View source: R/mixturetarget.R

Description

Create the posterior distribution of the parameters of a mixture of univariate gaussian distributions, with a fixed (known) number of components.

Usage

1
    createMixtureTarget(mixturesample, mixturesize, ncomponents, mixtureparameters)

Arguments

All the arguments are optional, since if none is given, a mixture distribution with 4 components will be created, as in Jasra, Holmes, Stephens, "MCMC and label switching problem in Bayesian mixture models", published in Statistical Science (2005).

mixturesample

Object of class "vector": data set to be used. If not provided, a synthetic data set is generated.

mixturesize

Object of class "numeric": represents the data set size if a data set is to be generated.

ncomponents

Object of class "numeric": represents the fixed number of components to be used.

mixtureparameters

Object of class "list": provides the parameters to be used if a data set has to be generated. The parameters include the number of components, the component weights, means and variances.

Value

The function returns an object of class target-class, with a name, a dimension, a function giving the log density, a function to generate sample from the distribution, parameters of the distribution, and a function to draw init points for the MCMC algorithms. The log density involves a likelihood and a prior, and the prior is as in Richardson and Green, "On Bayesian analysis of mixtures with an unknown number of components", published in JRSS B, 1997.

Author(s)

Luke Bornn <bornn@stat.harvard.edu>, Pierre E. Jacob <pierre.jacob.work@gmail.com>

References

Jasra, Holmes, Stephens, "MCMC and label switching problem in Bayesian mixture models", published in Statistical Science (2005). Richardson and Green, "On Bayesian analysis of mixtures with an unknown number of components", published in JRSS B, 1997.

See Also

target-class, createTrimodalTarget


PAWL documentation built on May 2, 2019, 5:58 a.m.