distcritmulti: Distance based validity criteria for large data sets

Description

Approximates average silhouette width or the Pearson version of Hubert's gamma criterion by hacking the dataset into pieces and averaging the subset-wise values, see Hennig and Liao (2013).

Usage

distcritmulti(x,clustering,part=NULL,ns=10,criterion="asw",
                    fun="dist",metric="euclidean",
                     count=FALSE,seed=NULL,...)

Arguments

cases times variables data matrix.

clustering

vector of integers indicating the clustering.

part

vector of integer subset sizes; sum should be smaller or equal to the number of cases of x. If NULL, subset sizes are chosen approximately equal.

integer. Number of subsets, only used if part==NULL.

criterion

"asw" or "pearsongamma", specifies whether the average silhouette width or the Pearson version of Hubert's gamma is computed.

fun

"dist" or "daisy", specifies which function is used for computing dissimilarities.

metric

passed on to dist (as argument method) or daisy to determine which dissimilarity is used.

count

logical. if TRUE, the subset number just processed is printed.

seed

integer, random seed. (If NULL, result depends on random numbers.)

...

further arguments to be passed on to dist or daisy.

Value

A list with components crit.overall,crit.sub,crit.sd,part.

crit.overall

value of criterion.

crit.sub

vector of subset-wise criterion values.

crit.sd

standard deviation of crit.sub, can be used to assess stability.

subsets

list of case indexes in subsets.

References

Halkidi, M., Batistakis, Y., Vazirgiannis, M. (2001) On Clustering Validation Techniques, Journal of Intelligent Information Systems, 17, 107-145.

Hennig, C. and Liao, T. (2013) How to find an appropriate clustering for mixed-type variables with application to socio-economic stratification, Journal of the Royal Statistical Society, Series C Applied Statistics, 62, 309-369.

Kaufman, L. and Rousseeuw, P.J. (1990). "Finding Groups in Data: An Introduction to Cluster Analysis". Wiley, New York.

Examples

Run this code

# NOT RUN {
    set.seed(20000)
    options(digits=3)
    face <- rFace(50,dMoNo=2,dNoEy=0,p=2)
    clustering <- as.integer(attr(face,"grouping"))
    distcritmulti(face,clustering,ns=3,seed=100000,criterion="pearsongamma")
# }