mcmcestimate: Calculate point estimators from MCMC samples

In simonsays1980/finmix: An R package for Bayesian estimation of finite mixture distributions

Description

Calling mcmcestimate() calculates the following point estimates from the MCMC samples:

• MAP: The maximum a posterior estimates are defined as the mode of the (joint) posterior density.

• BML: The Bayesian maximum likelihood estimator is based on the mixture log-likelihood function and defines the mode of this function.

• EAVG: The ergodic average is calculated as an average over the MCMC traces of component parameters and weights (in case of unknown parameters).

• IEAVG: The identified ergodic average is defined similar to the EAVG, however, in contrast to the latter it is based on re-labeled MCMC traces. This is especially important in case of random permutation during MCMC sampling as component parameters then have to be re-assigned to their (probably) correct component.

For a more detailed outlay of point estimators from Bayesian mixture model estimation, see Fr\"uhwirth-Schnatter (2006).

Usage

mcmcestimate(
  mcmcout,
  method = "kmeans",
  fdata = NULL,
  permOut = FALSE,
  opt_ctrl = list(max_iter = 200L)
)

Arguments

mcmcout	An mcmcoutput object containing the sampled parameters and informaiton about the finite mixture model.
method	A character defining the re-labeling method in case of a model with unknown indicators. For most distributions there exists only a single choice, namely "kmeans". For Poisson and Binomial distributions the re-labeling algorithms "Stephens1997a" and "Stephens1997b" can be chosen.
fdata	An fdata model containing the observations. Optional.
permOut	A logical indicating, if the permuted MCMC samples should be returned as well. Optional.
opt_ctrl	A list with an element max_iter controlling the number of iterations in case the "Stephens1997a" re-labeling algorithm is chosen.

Value

An mcmcest object containing the point estimates together with additional information about the underlying finite mixture model, MCMC sampling hyper-parameters and the data. In case permOut is set to TRUE, the output of this function is a named list with an mcmcest object containing parameter estimates and in addition an mcmcoutputperm object containing the permuted (re-labeled) MCMC samples.

See Also

• mcmcestfix for object storing the parameter estimates in case of fixed indicators

• mcmcestind for object storing the parameter estimates in case of unknown indicators

• mcmcoutputperm for classes storing re-labeled MCMC samples

simonsays1980/finmix documentation built on Dec. 23, 2021, 2:25 a.m.