NB.MClust {NB.MClust}

NB.MClust Function

Description

This function performs model-based clustering on positive integer or continuous data that follow Generalized Negative Binomial distribution.

Usage

NB.MClust(Count, K, ini.shift.mu = 0.01, ini.shift.theta = 0.01,
  tau0 = 10, rate = 0.9, bic = TRUE, iteration = 100)

Arguments

Count

Data matrix of discrete counts.This function groups rows of the data matrix.

K

Number of clusters or components specified. It can be a positive integer or a vector of positive integer.

ini.shift.mu

Initial value in EM algorithm for the shift between clusters in mean.

ini.shift.theta

Initial value in EM algorithm for the shift between clusters in dispersion.

tau0

Initial value of anealing rates in EM Algorithm. Default and suggested value is 10.

rate

Stochastic decreasing speed for anealing rate. Default and suggested value is 0.9

bic

Whether Bayesian Information should be computed when K is an integer. BIC is forced to be TRUE when K is a vector.

iteration

Maximum number of iterations in EM Algorithm, default at 50.

Value

parameters

Estimated parameters

$prior

Prior probability that a sample belongs to each cluster

$mu

Mean of each cluster

$theta

Dispersion of each cluster

$posterior

Posterior probability that a sample belongs to each cluster

cluster

Estimated cluster assignment

BIC

Value of Bayesian Information

K

Optional or estimated number of clusters, if input K is a vector

Examples

# Example:

data("Simulated_Count") # A 50x100 integer data frame.

m1=NB.MClust(Simulated_Count,K=2:5)
cluster=m1$cluster #Estimated cluster assignment
k_hat=m1$K  #Estimated optimal K