Package 'PseudoVoigtMixt'

Title: Multivariate Pseudo-Voigt Mixture Models
Description: Fits multivariate pseudo-Voigt mixture models for model-based clustering and outlier detection. The model combines multivariate Gaussian and Cauchy distributions within each cluster to accommodate heavy-tailed observations and decompose the data into a high-density region and a low-density remainder, with outliers more likely to arise from the latter.
Authors: Babak Fathollahi Dehkordi [aut, cre], Jeffery Andrews [aut, ths]
Maintainer: Babak Fathollahi Dehkordi <[email protected]>
License: MIT + file LICENSE
Version: 0.1.0
Built: 2026-10-09 09:17:50 UTC
Source: https://github.com/cran/PseudoVoigtMixt

Help Index

Fit Multivariate Pseudo-Voigt Mixture Models Estimates MPVM models over a range of numbers of components and selects the best-fitting model according to the Bayesian information criterion (BIC).

Description

Fit Multivariate Pseudo-Voigt Mixture Models Estimates MPVM models over a range of numbers of components and selects the best-fitting model according to the Bayesian information criterion (BIC).

Usage

MPVmixt(
  x,
  G_range,
  max_iter = 100,
  init_method = c("modified_tkmeans", "manual"),
  tol = 1e-08,
  alpha_trim = 0.3,
  nstart = 1000,
  verbose = FALSE,
  seed = NULL,
  shared_scale = TRUE,
  Z_init = NULL,
  V_init = NULL
)

Arguments

x

Numeric matrix or data frame containing the observations. Rows correspond to observations and columns to variables.
G_range

Positive integer or vector of positive integers specifying the number of mixture components to fit.
max_iter

Maximum number of EM iterations. Default is 100.
init_method

Initialization method. Either '"modified_tkmeans"' or '"manual"'.
tol

Positive convergence tolerance for the log-likelihood.
alpha_trim

Trimming proportion used by the modified trimmed k-means initialization.
nstart

Number of random starts used by the modified trimmed k-means initialization.
verbose

Logical; if 'TRUE', iteration information is printed.
seed

Optional random seed for reproducibility.
shared_scale

Logical; if 'TRUE', the Gaussian and Cauchy sub-components share the same scale matrix within each cluster.
Z_init

Optional initial cluster-membership matrix used when 'init_method = "manual"'.
V_init

Optional initial Gaussian/Cauchy sub-component-membership matrix used when 'init_method = "manual"'.

Value

A list containing:

all

Results for each value in 'G_range'.

best

The model corresponding to the minimum BIC, including clustering and outlier-detection results.

bic

Numeric vector of BIC values corresponding to 'G_range'.

#' @examples sim <- rmpvm( n = 100, pi = c(0.5, 0.5), alpha = c(0.8, 0.8), mu = list( c(0, 0), c(4, 4) ), Sigma = list( diag(2), diag(2) ), Gamma = list( diag(2), diag(2) ), seed = 123 )

fit <- MPVmixt( x = sim$data, G_range = 2, max_iter = 20, nstart = 20, seed = 123 )

fit$best$G

Generate Data from a Multivariate Pseudo-Voigt Mixture Model

Description

Generate Data from a Multivariate Pseudo-Voigt Mixture Model

Usage

rmpvm(n, pi, alpha, mu, Sigma, Gamma, seed = NULL)

Arguments

n

Number of observations to generate.
pi

Numeric vector of mixing proportions.
alpha

Numeric vector of Gaussian subcomponent proportions.
mu

List of component location vectors.
Sigma

List of Gaussian covariance matrices.
Gamma

List of Cauchy scale matrices.
seed

Optional random seed for reproducibility.

Value

A list containing the generated data matrix, latent cluster labels, latent Gaussian/Cauchy indicators, and corresponding factor labels.

#' @examples sim <- rmpvm( n = 100, pi = c(0.5, 0.5), alpha = c(0.8, 0.8), mu = list( c(0, 0), c(4, 4) ), Sigma = list( diag(2), diag(2) ), Gamma = list( diag(2), diag(2) ), seed = 123 )

head(sim$data)