ecm: Model estimation through ECM

In ben-j-barlow/rcbmm: Regularized Copula-Based Mixture Model: Estimation and Model Selection

Description

Implements the expectation-conditional-maximization algorithm for a regularized copula-based mixture model given initial parameter values, starting with the expectation step.

Usage

ecm(
  x,
  K,
  lambda,
  start = NULL,
  margins,
  trace = TRUE,
  maxit = 1000,
  epsilon = 1e-06,
  dist_mat = NULL
)

Arguments

x	A numeric matrix or data frame of observations. Rows correspond to observations and columns correspond to variables.
K	An integer specifying the number of components for which a regularized copula-based mixture model should be fitted.
lambda	A numeric value indicating the value of the tuning parameter for regularization.
start	A list providing the starting values for ECM. The list is produced by fit.rcbmm.
margins	A character vector specifying the marginal distributions of the components in the mixture. The vector must have a length equal to the number of columns in x. Each element must be equal to "norm", "beta" or "gamma".
trace	A logical value indicating if an update regarding the algorithm's progress should be displayed after each iteration.
maxit	A numeric value indicating the maximal number of ECM iterations.
epsilon	A numeric value specifying the tolerance associated with determining when convergence of the ECM algorithm has been achieved.
dist_mat	An object of type dist for calculating silhouette values.

Value

• KThe number of mixture components.

• lambdaThe value of the tuning parameter.

• zA numeric matrix representing the posterior probabilities of membership of the observations after the expectation step of the last iteration of the ECM algorithm. Columns are associated with a mixture component and rows are associated with observations.

• clustersA classification vector indicating the associated cluster of each observation. The classification corresponds to z.

• loglikA numeric vector displaying the penalized log-likelihood after each iteration of ECM.

• param_numberThe number of independent parameters associated with the model.

• BICThe BIC value of the model. Computed using the unpenalized log-likelihood after the last iteration of the ECM algorithm.

• mixing_probsThe mixing proportions associated with the model.

• mvdcA list of objects of class mvdc. Each element of the list corresponds to a mixture component.

• transformationA character value indicating the transformation when identifying starting values. The value is NULL unless lambda=0. See initialize.ecm.

• marginal_paramA list containing the marginal parameters of the model as estimated by ECM. Each element corresponds to a mixture component.

• copula_paramA list containing the copula parameters of the model as estimated by ECM. Each element corresponds to a mixture component.

• copula_param_anglesA list containing the copula parameters of the model re-expressed as angles.

• silhouetteSee information regarding silhouette package. Add reference here.

ben-j-barlow/rcbmm documentation built on Feb. 12, 2021, 9:14 a.m.