disclapmix (version 1.7.5)

disclapmix_adaptive: Adaptive fitting

Description

A wrapper around `disclapmix_robust()` that instead of fitting one model for a given number of clusters, fits models until the best model (lowest marginal BIC) is in the interior (with margin `M`) of all number of clusters tried.

Usage

disclapmix_adaptive(
  x,
  label = "DL",
  margin = 5L,
  criteria = "BIC",
  init_y_generator = NULL,
  init_v_generator = NULL,
  ...
)

Value

A list of all `disclapmix` fits

Arguments

x

Dataset.

label

A label

margin

Fit models until there is at least this margin

criteria

The slot to chose the best model from (BIC/AIC/AICc)

init_y_generator

Function taking the number of clusters as input and returns `init_y` values

init_v_generator

Function taking the number of clusters as input and returns `init_v` values

...

Passed on to `disclapmix_robust()` (and further to `disclapmix()`)

Details

E.g., the best model has 3 clusters and the margin `M = 5`, then this function ensures that models with 1, 2, ..., 3+5 = 8 clusters are fitted. If e.g. then 7 is better than 3, then it continues such that also models with up to 7+5 = 12 clusters are fitted.

Note that models with 1-5 clusters are always fitted.

Examples

data(danes)
db <- as.matrix(danes[rep(1:nrow(danes), danes$n), 1:(ncol(danes)-1)])
fits <- disclapmix_adaptive(db, margin = 5L)
fits
BICs <- sapply(fits, function(x) x$BIC_marginal)
BICs
ks <- sapply(fits, function(x) nrow(x$y)) # Always same as seq_along(fits)
ks
max_k <- max(ks)
best_k <- which.min(BICs)
max_k
best_k
max_k - best_k # = margin = 5
plot(ks, BICs, type = "b")

fits_clara <- disclapmix_adaptive(db, margin = 5L, init_y_method = "clara")