mvnmix function

Estimate mixture latent variable model

mvnmix(
  data,
  k = 2,
  theta,
  steps = 500,
  tol = 1e-16,
  lambda = 0,
  mu = NULL,
  silent = TRUE,
  extra = FALSE,
  n.start = 1,
  init = "kmpp",
  ...
)

Arguments

• data: data.frame
• k: Number of mixture components
• theta: Optional starting values
• steps: Maximum number of iterations
• tol: Convergence tolerance of EM algorithm
• lambda: Regularisation parameter. Added to diagonal of covariance matrix (to avoid singularities)
• mu: Initial centres (if unspecified random centres will be chosen)
• silent: Turn on/off output messages
• extra: Extra debug information
• n.start: Number of restarts
• init: Function to choose initial centres
• ...: Additional arguments parsed to lower-level functions

Returns

A mixture object

Details

Estimate parameters in a mixture of latent variable models via the EM algorithm.

Examples

data(faithful)
set.seed(1)
M1 <- mvnmix(faithful[,"waiting",drop=FALSE],k=2)
M2 <- mvnmix(faithful,k=2)
if (interactive()) {
    par(mfrow=c(2,1))
    plot(M1,col=c("orange","blue"),ylim=c(0,0.05))
    plot(M2,col=c("orange","blue"))
}

See Also

mixture

Author(s)

Klaus K. Holst