CemCOVar {cemco}

Fit CemCO algorithm with covariates effects on cluster centroids and covariance matrices.

Description

Model-based clustering based on parameterized finite Gaussian mixture models with covariates effects on the distribution means and the distribution covariance matrices. Models are estimated by an EM algorithm.

Usage

CemCOVar(data, y, G, y_cov, max_iter=100, n_start=20, cores=4)

Arguments

data

A numeric vector, matrix, or data frame of observations. Non-numerical values should be converted to integer or float (e.g. dummies). If matrix or data frame, rows and columns correspond to observations (n) and variables (P).

y

numeric matrix of data to use as covariates. Non-numerical values should be converted to integer or float (e.g. dummies).

G

An integer specifying the numbers of mixture components (clusters)

y_cov

numeric matrix of data to use as covariates for the covariance effect. Non-numerical values should be converted to integer or float (e.g. dummies)

max_iter

maximum number of iterations of the EM optimization (default value equals to 100)

n_start

how many random sets should be chosen? (default value equals to 20)

cores

number of cores for EM optimization (default value equals to 4)

Details

This function optimizes the log likelihood of the CemCO algorithm with covariates effects on cluster centroids and covariance matrices using a implementation of the EM algorithm. The covariates associated with the distributions means and with the distributions covariance matrices do not need to be the same.

Value

The function output is a list

fitted parameters

The estimated parameters of the CemCO algorithm, including clusters centroids, covariance matrix, covariate effects of each cluster and a priori probability of each cluster.

log likelihood

The optimal log likelihood estimated by the model

Author(s)

Relvas, C. & Fujita, A.

References

Stage I non-small cell lung cancer stratification by using a model-based clustering algorithm with covariates, Relvas et al.

Examples

set.seed(42)
X = cbind(rnorm(20), rnorm(20))
Y = cbind(rnorm(20), rnorm(20))
K = 2


fit <- CemCOVar(X, Y, K, Y[,1], max_iter=5, n_start=1, cores=1)
params <- fit[[1]]  ## fitted parameters
ll <- fit[[2]]  ## log likelihood