R: Marginal (univariate) predictive cumulative distribution...

NMixPredCDFMarg {mixAK}

R Documentation

Marginal (univariate) predictive cumulative distribution function

Description

This function serves as an inference tool for the MCMC output obtained using the function NMixMCMC. It computes estimated posterior predictive cumulative distribution function for each margin.

Usage

NMixPredCDFMarg(x, ...)

## Default S3 method:
NMixPredCDFMarg(x, scale, K, w, mu, Li, Krandom=TRUE, ...)

## S3 method for class 'NMixMCMC'
NMixPredCDFMarg(x, grid, lgrid=500, scaled=FALSE, ...)

## S3 method for class 'GLMM_MCMC'
NMixPredCDFMarg(x, grid, lgrid=500, scaled=FALSE, ...)

Arguments

`x`	an object of class `NMixMCMC` for `NMixPredCDFMarg.NMixMCMC` function. An object of class `GLMM_MCMC` for `NMixPredCDFMarg.GLMM_MCMC` function. A list with the grid values (see below) for `NMixPredCDFMarg.default` function.
`scale`	a two component list giving the `shift` and the `scale`.
`K`	either a number (when `Krandom=FALSE`) or a numeric vector with the chain for the number of mixture components.
`w`	a numeric vector with the chain for the mixture weights.
`mu`	a numeric vector with the chain for the mixture means.
`Li`	a numeric vector with the chain for the mixture inverse variances (lower triangles only).
`Krandom`	a logical value which indicates whether the number of mixture components changes from one iteration to another.
`grid`	a numeric vector or a list with the grid values in which the predictive CDF should be evaluated. If `x$dim` is 1 then `grid` may be a numeric vector. If `x$dim` is higher than then `grid` must be a list with numeric vectors as components giving the grids for each margin. If `grid` is not specified, it is created automatically using the information from the posterior summary statistics stored in `x`.
`lgrid`	a length of the grid used to create the `grid` if that is not specified.
`scaled`	if `TRUE`, the CDF of shifted and scaled data is summarized. The shift and scale vector are taken from the `scale` component of the object `x`.
`...`	optional additional arguments.

Value

An object of class NMixPredCDFMarg which has the following components:

`x`	a list with the grid values for each margin. The components of the list are named `x1`, ... or take names from `grid` argument.
`freqK`	frequency table for the values of `K` (numbers of mixture components) in the MCMC chain.
`propK`	proportions derived from `freqK`.
`MCMC.length`	the length of the MCMC used to compute the predictive cdf's.
`cdf`	a list with the computed predictive CDF's for each margin. The components of the list are named `1`, ..., i.e., `cdf[[1]]=cdf[["1"]]` is the predictive cdf for margin 1 etc.
`cdfK`	a list with the computed predictive CDF's for each margin, conditioned further by `K`. The components of the list are named `1`, .... That is, `cdf[[1]][[1]]` `=` `cdf[["1"]][[1]]` is the predictive CDF for margin 1 conditioned by `K=1`, `cdf[[1]][[2]]` `=` `cdf[["1"]][[2]]` is the predictive CDF for margin 1 conditioned by `K=2` etc. Note that `cdfK` provides some additional information only when `Krandom` `=` `TRUE` or when `x` results from the `NMixMCMC` call to the reversible jump MCMC.

There is also a plot method implemented for the resulting object.

Author(s)

Arnošt Komárek arnost.komarek@mff.cuni.cz

References

Komárek, A. (2009). A new R package for Bayesian estimation of multivariate normal mixtures allowing for selection of the number of components and interval-censored data. Computational Statistics and Data Analysis, 53(12), 3932–3947.