bqs_select: Select Ranked Cluster Solutions by Quadratic Score

In qcluster: Clustering via Quadratic Scoring

Select Ranked Cluster Solutions by Quadratic Score

Description

Select solutions from a bqs object based on specified rank and type of score.

Usage

bqs_select(
  bqs_sol,
  rank = 1,
  type = "smooth",
  rankby = NA,
  boot_na_share = 0.25
)

Arguments

bqs_sol	An object of class bqs containing the clustering solutions to be selected.
rank	An integer >0 specifying the rank of the solution to select. Default is 1.
type	A character string specifying the type of Quadratic Score. Possible values are "hard", "smooth", "oob_hard", "oob_smooth". Default is "smooth".
rankby	A character string specifying the criteria used to rank solutions in bqs. Possible values are "lq", "mean", "1se", or NA (default). See Details.
boot_na_share	A numeric value between [0, 1]. Clustering solutions in bqs_sol with a share of NA bootstrap estimates are excluded from ranking. Default is 0.25.

Details

Even if the bqs_sol object is not pre-ranked, the user may specify a ranking criterion to rank clustering solutions dynamically using the rankby argument; this does not modify the original bqs_sol object. In these instances, the user can also specify boot_na_share as in bqs_rank to exclude solutions based on the proportion of unsuccessful bootstrap estimations. If rankby=NA, the bqs_sol must be pre-ranked.

Selected solutions are always re-estimated on the full dataset before being returned. Therefore, for stochastic clustering methods, repeated calls may return different fitted solutions unless the user controls reproducibility, e.g. via set.seed().

Value

A named list of all clustering solutions achieving a type score of rank rank when ranked according to rankby criterion, or NULL if no such solution is available in the bqs_sol object. List names correspond to methods' names in bqs_sol$methodset. Each named entry contains the corresponding method re-estimated on bqs_sol$data$data. If the full-data refit fails, the corresponding entry is an object of class bqs_select_error containing the failure status and message. If the requested rank exceeds the largest available rank, the worst available rank is returned instead. If the requested rank is within range but absent because of rank gaps, NULL is returned.

See Also

bqs, bqs_rank

Examples

# Load data and set seed
set.seed(123)
data("banknote")
dat <- banknote[-1]

# set up kmeans, see help('mset_kmeans')
KM <- mset_kmeans(K = 2:5)

# set up Gaussian model-based clustering via gmix()
GMIX <- mset_gmix(K = 2:5, erc = c(1, 50, 100))

# combine tuned methods
mlist <- mbind(KM, GMIX)

# perform bootstrap
# set 'ncores' to the number of available physical cores
res <- bqs(dat, mlist, B = 20, type = "both", rankby = NA, ncores = 1,
           oob = TRUE, savescores = TRUE, saveparams = FALSE)

# Methods are not ranked; this will raise an error
try(bqs_select(res, rank = 1))

# Rank method dynamically
ranked_res <- bqs_select(res, rank = 2, rankby = "lq",
                         boot_na_share = 0.25)
names(ranked_res)

qcluster documentation built on June 5, 2026, 5:07 p.m.