mcmcarray-object: Objects for representing MCMC output.

In biips/rbiips: Bayesian Inference with Interacting Particle Systems

Description

A mcmcarray object is returned by the biips_pimh_samples or biips_pmmh_samples functions to represent MCMC output of a given variable.

A mcmcarray.list object is a named list of mcmcarray objects for different monitored variables.

The methods apply identically to mcmcarray or mcmcarray.list objects and return a named list with the same named members as the input object.

Usage

mcmcarray(data = NA, dim = length(data), dimnames = NULL,
  iteration = length(dim), chain = NA, name = "mcmcarray", lower = NULL,
  upper = NULL)

is.mcmcarray(object)

is.mcmcarray.list(object)

## S3 method for class 'mcmcarray'
biips_summary(object, probs = c(), order = ifelse(mode,
  0, 1), mode = all(object == as.integer(object)), ...)

## S3 method for class 'mcmcarray.list'
biips_summary(object, ...)

## S3 method for class 'mcmcarray'
biips_table(x, ...)

## S3 method for class 'mcmcarray'
biips_density(x, bw = "nrd0", ...)

biips_hist(x, ...)

## S3 method for class 'mcmcarray'
biips_hist(x, main = NULL, xlab = NULL, ...)

## S3 method for class 'mcmcarray.list'
biips_table(x, ...)

## S3 method for class 'mcmcarray.list'
biips_density(x, bw = "nrd0", ...)

## S3 method for class 'mcmcarray.list'
biips_hist(x, main = NULL, xlab = NULL, ...)

## S3 method for class 'mcmcarray'
summary(object, ...)

## S3 method for class 'mcmcarray.list'
summary(object, ...)

## S3 method for class 'mcmcarray'
density(x, ...)

## S3 method for class 'mcmcarray.list'
density(x, ...)

## S3 method for class 'mcmcarray'
hist(x, ...)

## S3 method for class 'mcmcarray.list'
hist(x, ...)

Arguments

data	numerical vector
dim	vector of integers. dimension of the array
dimnames	character vector
iteration	integer. index of the dimension corresponding to iterations of the MCMC.
chain	integer. index of the dimension corresponding to chain of the MCMC.
name	string. variable name
lower	vector of integers. variable lower bound
upper	vector of integers. variable upper bound
object, x	a mcmcarray or mcmcarray.list object.
probs	vector of reals. probability levels in ]0,1[ for quantiles. (default = c())
order	integer. Moment statistics of order below or equal to order are returned. (default = 0 if all the components are discrete variables and 1 otherwise)
mode	logical. Activate computation of the mode, i.e. the most frequent value among the particles. (default = TRUE if all the components are discrete variables and FALSE otherwise)
...	additional arguments to be passed to the default methods. See density, hist, table
bw	either a real with the smoothing bandwidth to be used or a string giving a rule to choose the bandwidth. See bw.nrd. (default='nrd0')
main, xlab	plotting parameters with useful defaults.

Value

The methods apply identically to mcmcarray or mcmcarray.list objects and return a named list with the same named members as the input object.

The mcmcarray function returns an object of class mcmcarray.

The function is.mcmcarray returns TRUE if the object is of class mcmcarray.

The function is.mcmcarray.list returns TRUE if the object is of class mcmcarray.list.

The method biips_summary returns univariate marginal statistics. The output innermost members are objects of class summary.mcmcarray, i.e. lists with members:

mean	mean, if order>=1.
var	variance, if order>=2.
skew	skewness, if order>=3.
kurt	kurtosis, if order>=4.
probs	vector of quantile probabilities.
quant	list of quantile values, if probs is not empty.
mode	most frequent values for discrete components.

The method biips_table returns univariate marginal frequency tables or probability mass estimates of discrete variables. The output innermost members are objects of class table.mcmcarray.

The method biips_density returns univariate marginal kernel density estimates. The output innermost members are objects of class density.mcmcarray.

The method summary is an alias for biips_summary.

The method density is an alias for biips_density.

The method hist is an alias for biips_hist.

See Also

density

Examples

modelfile <- system.file('extdata', 'hmm.bug', package = 'rbiips')
stopifnot(nchar(modelfile) > 0)
cat(readLines(modelfile), sep = '\n')

#' # PIMH algorithm
data <- list(tmax = 10, p = c(.5, .5), logtau_true = log(1), logtau = log(1))
model <- biips_model(modelfile, data, sample_data = TRUE)

n_part <- 50
obj_pimh <- biips_pimh_init(model, c('x', 'c[2:10]'))  # Initialize
out_pimh_burn <- biips_pimh_update(obj_pimh, 100, n_part)  # Burn-in
out_pimh <- biips_pimh_samples(obj_pimh, 100, n_part)  # Samples

#' Manipulate `mcmcarray.list` object
is.mcmcarray.list(out_pimh)
names(out_pimh)
out_pimh
biips_summary(out_pimh)

#' Manipulate `mcmcarray` object
is.mcmcarray(out_pimh$x)
out_pimh$x
summ_pimh_x <- biips_summary(out_pimh$x, order = 2, probs = c(0.025, 0.975))
summ_pimh_x
dens_pimh_x <- biips_density(out_pimh$x)
par(mfrow = c(2, 2))
plot(dens_pimh_x)
par(mfrow = c(2, 2))
biips_hist(out_pimh$x)

is.mcmcarray(out_pimh[['c[2:10]']])
out_pimh[['c[2:10]']]
summ_pimh_c <- biips_summary(out_pimh[['c[2:10]']])
summ_pimh_c
table_pimh_c <- biips_table(out_pimh[['c[2:10]']])
par(mfrow = c(2, 2))
plot(table_pimh_c)

#' # PMMH algorithm
data <- list(tmax = 10, p = c(.5, .5), logtau_true = log(1))
model <- biips_model(modelfile, data)

n_part <- 50
obj_pmmh <- biips_pmmh_init(model, 'logtau', latent_names = c('x', 'c[2:10]'),
                            inits = list(logtau = -2))  # Initialize
out_pmmh_burn <- biips_pmmh_update(obj_pmmh, 100, n_part)  # Burn-in
out_pmmh <- biips_pmmh_samples(obj_pmmh, 100, n_part, thin = 1)  # Samples

#' Manipulate `mcmcarray.list` object
is.mcmcarray.list(out_pmmh)
names(out_pmmh)
out_pmmh
biips_summary(out_pmmh)

#' Manipulate `mcmcarray` object
is.mcmcarray(out_pmmh$logtau)
out_pmmh$logtau
summ_pmmh_lt <- biips_summary(out_pmmh$logtau, order = 2, probs = c(0.025, 0.975))
dens_pmmh_lt <- biips_density(out_pmmh$logtau)
par(mfrow = c(2, 1))
plot(dens_pmmh_lt)
biips_hist(out_pmmh$logtau)

is.mcmcarray(out_pmmh$x)
out_pmmh$x
summ_pmmh_x <- biips_summary(out_pmmh$x, order = 2, probs = c(0.025, 0.975))
dens_pmmh_x <- biips_density(out_pmmh$x)
par(mfrow = c(2, 2))
plot(dens_pmmh_x)
par(mfrow = c(2, 2))
biips_hist(out_pmmh$x)

is.mcmcarray(out_pmmh[['c[2:10]']])
out_pmmh[['c[2:10]']]
summ_pmmh_c <- biips_summary(out_pmmh[['c[2:10]']])
table_pmmh_c <- biips_table(out_pmmh[['c[2:10]']])
par(mfrow = c(2, 2))
plot(table_pmmh_c)

biips/rbiips documentation built on Nov. 28, 2020, 2:12 p.m.