Update BIC values for parameterized Gaussian mixture models

Update the BIC (Bayesian Information Criterion) for parameterized Gaussian mixture models by taking the best from BIC results as returned by mclustBIC.

Usage

mclustBICupdate(BIC, ...)

Arguments

BIC

Object of class 'mclustBIC' containing the BIC values as returned by a call to mclustBIC.

...

Further objects of class 'mclustBIC' to be merged.

Value

An object of class 'mclustBIC' containing the best values obtained from merging the input arguments. Attributes are also updated according to the best BIC found, so calling Mclust on the resulting ouput will return the corresponding best model (see example).

See also

mclustBIC, Mclust.

Examples

# \donttest{
data(galaxies, package = "MASS") 
galaxies <- galaxies / 1000

# use several random starting points
BIC <- NULL
for(j in 1:100)
{
  rBIC <- mclustBIC(galaxies, verbose = FALSE,
                    initialization = list(hcPairs = hcRandomPairs(galaxies)))
  BIC <- mclustBICupdate(BIC, rBIC)
}
pickBIC(BIC)
#>       V,3       V,5       V,4 
#> -441.6122 -441.8364 -443.3891 
plot(BIC)


mod <- Mclust(galaxies, x = BIC)
summary(mod)
#> ---------------------------------------------------- 
#> Gaussian finite mixture model fitted by EM algorithm 
#> ---------------------------------------------------- 
#> 
#> Mclust V (univariate, unequal variance) model with 3 components: 
#> 
#>  log-likelihood  n df       BIC       ICL
#>       -203.1792 82  8 -441.6122 -441.6126
#> 
#> Clustering table:
#>  1  2  3 
#> 72  7  3 
# }