MCMC sampling from a given dataset

MCMC sampling from a given dataset

Usage

mcmcSampling(
  dataset = NULL,
  dimensions = list(""),
  densityFunction = alwaysOne,
  proposalFunction = addHighDimGaussian(dim = length(dimensions)),
  n.sample.points = 0,
  burnIn = 1000,
  verbose = TRUE,
  covariance.correction = 1,
  max.burnin.cycles = 50,
  engine = c("auto", "R", "cpp")
)

Arguments

dataset

sf dataframe from which the points are sampled

dimensions

string vector containing the dimensions that should be included in the random walk

densityFunction

Function that can take a point given as a numeric vector as input and returns the target density at that location.

proposalFunction

Function that can take a point given as a numeric vector and a covariance adjuster as input and returns a new proposed point as a numeric vector.

n.sample.points

Number of points to be sampled

burnIn

Integer, number of Robbins-Monro burn-in adaptation steps performed before sampling. During each step the proposal scale is adjusted toward target acceptance 0.234 (Roberts/Rosenthal 2009). Set to 0 to skip adaptation and start sampling immediately at the user-supplied covariance.correction.

verbose

Boolean to toggle progress updates

covariance.correction

Integer, initial value of the covariance correction.

max.burnin.cycles

Deprecated. Retained for backwards compatibility; ignored by the current Robbins-Monro burn-in.

engine

One of "auto" (default), "R", or "cpp". "auto" picks the C++ inner loop when both densityFunction and proposalFunction are built by mclustDensityFunction() and addHighDimGaussian() (they carry the required rcpp_spec attribute) and falls back to the R loop otherwise. "cpp" forces the C++ path and errors if a custom closure is supplied. "R" forces the pure-R reference loop.

Value

A data.frame containing the sampled points with dimension columns and a density column