KcopClust {Kcop}

Nonparametric clustering of multivariate populations with arbitrary sizes

Description

This function performs the data driven clustering procedure to cluster K multivariate populations of arbitrary sizes into N subgroups characterized by a common dependence structure where the number N of clusters is unknow and will be automatically chosen by our approach. The method is adapted to paired population and can be used with panel data. See the paper at the following arXiv weblink: https://arxiv.org/abs/2211.06338 for further information.

Usage

KcopClust(Kdata, dn = 3, paired = FALSE, alpha = 0.05)

Arguments

Kdata

A list of the K dataframe or matrix

dn

Number of copulas coefficients considered

paired

A logical indicating whether to consider the datas as paired

alpha

The significance level used in our decision rule.

Value

A list with three elements: the number of identified clusters; 2) the cluster affiliation; 3) the discrepancy matrix. the numbers in the clusters refer to the population indexes of the data list

Author(s)

Yves I. Ngounou Bakam and Denys Pommeret

Examples

## simulation of 5 three-dimensional populations of different sizes
Packages <- c("copula","gtools","dplyr", "orthopolynom", "stats")
lapply(Packages, library, character.only = TRUE) # if necessary
set.seed(2022)
dat1<-rCopula(50, copula = gumbelCopula(param=6,dim = 2))
dat2<-rCopula(60, copula = claytonCopula(param=0.4,dim = 2))
dat3<-rCopula(55, copula = claytonCopula(param=0.4,dim = 2))
## Form a list of data
Kdata<-list(data1=dat1,data2=dat2,data3=dat3)
## Applying the clustering
KcopClust(Kdata = Kdata)