RegClust (version 1.0)

cluster.reg: Clustering analysis of regression coefficients.

Description

This package performs clustering on regression coefficients using the methods of clustering through linear regression models (CLM) (Qin and Self 2006). Maximum likelihood approach is used to infer the parameters for each cluster. Bayesian information criterion (BIC) combined with Bootstrapped maximum volume (BMV) criterion are used to determine the number of clusters. The Bootstrap method is used to estimate the uncertainty on the number of clusters.

Usage

cluster.reg(Y, X, loop = 1000)

Arguments

Y

An n x k data matrix for n number of observations and k dependent variables.

X

An n x m data matrix for n number of observations and m covariates of interest. Restriction of m < n.

loop

A numeric scalar identfying the number of iterations for the bootstrappng process. Default is 1000.

Value

cluster

A numeric vector of length k identifying which cluster the respective dependent variables belong to.

parm

Dataframe containing estimates of regression coefficients, &sigma^2;, and π for each cluster, where &sigma^2; is the variance in the random error, and π is the probability that a variable is in a cluster.

likelihood

Likelihood at final iteration or at convergence.

BIC

Bayesian information criterion at final interation or at convergence.

References

Qin, Li-Xuan, and Steven G. Self. The clustering of regression models method with applications in gene expression data. Biometrics 62.2 (2006): 526-533.

Examples

## Not run: 
# beta0=1
# beta1=0.02
# beta2=-0.04
# beta3=0.1
# set.seed(1234)
# sim=200
# age=runif(sim, min=18, max=70)
# Rerror=rnorm(sim,-3,3)
# 
# y1=matrix(NA,sim,4,dimnames = list(NULL,c("y1", "y2", "y3", "y4")))
# for (g in 1:4){
# set.seed(1234)
# y1[,g]=beta0+beta1*runif(sim, min=18, max=70)+beta1*rnorm(sim,-3,3)
# set.seed(1134+g)
# y1[,g]=y1[,g]+rnorm(sim,0,1)
# }
# y2=matrix(NA,sim,5,dimnames = list(NULL,c("y5", "y6", "y7","y8","y9")))
# for (g in 1:5){
# set.seed(1234)
# y2[,g]=beta0+beta2*runif(sim, min=18, max=70)+beta2*rnorm(sim,-3,3)
# set.seed(2234+g)
# y2[,g]=y2[,g]+rnorm(sim,0,0.5^0.5)
# }
# y3=matrix(NA,sim,6,dimnames = list(NULL,c("y10", "y11", "y12","y13","y14","y15")))
# for (g in 1:6){
# set.seed(1234)
# y3[,g]=beta0+beta3*runif(sim, min=18, max=70)+beta3*rnorm(sim,-3,3)
# set.seed(3334+g)
# y3[,g]=y3[,g]+rnorm(sim,0,1)
# }
# X=data.frame(round(cbind(Rerror=Rerror,age=age),2))
# Y=data.frame(round(cbind(y1,y2,y3),2))
# 
# run<-cluster.reg(Y,X)
# run
# ## End(Not run)