sim.cluster.data: Simulated data for testing the BNPMIXcluster package

Description

Simulated values for three continuos variables under the existence of three clusters.

The data consists of a three-variate Normal distribution with different mean and covariance matrix between clusters.

This can be assumed either as continuos data to be clustered Y=(Y_1,Y_2,Y_3); or also can be used as the underlying latent data that can be transformed into observable variables Y_i=f(Z_i), which can be continuos or categorical.

Usage

sim.cluster.data

Arguments

Format

A data frame with 100 rows and 4 variables.

cluster: Indicates the cluster for each row
Z1,Z2,Z3: Continuos values coming from a multivariate normal distribution, given the cluster

Details

A data frame with 100 rows and 4 variables.

Examples

Run this code

# NOT RUN {
### Visualizing the simulated data for clustering ###

require(scatterplot3d)

cluster_color <- c(rgb(1,0,0,alpha = 0.5),
                   rgb(0,0,1,alpha = 0.5),
                   rgb(0,0.5,0,alpha = 0.5))
cluster_color <- cluster_color[sim.cluster.data$cluster]
cluster_pch <- c(19,15,17)[sim.cluster.data$cluster]
par(mfrow=c(2,2))
par(mar=c(4,5,2,2))

scatterplot3d::scatterplot3d(x=sim.cluster.data$Z1,y = sim.cluster.data$Z2, z=sim.cluster.data$Z3,
              color=cluster_color,pch=cluster_pch,
              xlab="Z1",ylab="Z2",zlab="Z3",
              main="Simulated data in 3 clusters"
              )
par(mar=c(4,5,2,2))
plot(sim.cluster.data[,c("Z2","Z3")],col=cluster_color,pch=cluster_pch,xlab="Z2",ylab="Z3")
par(mar=c(4,5,2,2))
plot(sim.cluster.data[,c("Z1","Z3")],col=cluster_color,pch=cluster_pch,xlab="Z1",ylab="Z3")
par(mar=c(4,5,2,2))
plot(sim.cluster.data[,c("Z1","Z2")],col=cluster_color,pch=cluster_pch,xlab="Z1",ylab="Z2")


### Code to generate the simulated data from scratch ###

require(MASS)
set.seed(0)

n.sim <- 100
n.cluster <- 3
p <- 3

mu_sim.cluster.data.Z <- matrix( c(2,2,5,
                              6,4,2,
                              1,6,2) ,
                           nrow=n.cluster, ncol=p, byrow=TRUE)

sigma_sim.cluster.data.Z <- array(dim=c(3,3,3))
sigma_sim.cluster.data.Z[,,1] <- diag(3)
sigma_sim.cluster.data.Z[,,2] <- matrix( c(0.1,0,0,
                                      0,2,0,
                                      0,0,0.1) ,
                           nrow=n.cluster, ncol=p, byrow=TRUE)

sigma_sim.cluster.data.Z[,,3] <- matrix( c(2,0,0,
                                      0,0.1,0,
                                      0,0,0.1) ,
                           nrow=n.cluster, ncol=p, byrow=TRUE)

sim.cluster.data <- data.frame(cluster=sample(x=1:n.cluster,size=n.sim,replace=TRUE))

sim.cluster.data.Z <- matrix(NA,nrow=n.sim,ncol=p)

for(i in unique(sim.cluster.data$cluster)) {
     sim.cluster.data.Z[sim.cluster.data[,1]==i,] <- MASS::mvrnorm(
                                                   n=sum(sim.cluster.data[,1]==i),
                                                   mu=mu_sim.cluster.data.Z[i,],
                                                   Sigma=sigma_sim.cluster.data.Z[,,i]
                                                   )
}
colnames(sim.cluster.data.Z) <- paste("Z",1:ncol(sim.cluster.data.Z),sep="")
sim.cluster.data <- cbind(sim.cluster.data,sim.cluster.data.Z)

# }