ClustMMDD (version 1.0.4)

em.cluster.R: Compute estimates of the parameters by Expectation and Maximization algorithm.

Description

Compute an approximation of the maximum likelihood estimates of parameters using Expectation and Maximization (EM) algorithm. A maximum a posteriori classification is then derived from the estimated set of parameters.

Usage

em.cluster.R(xdata, K, S, ploidy = 1, emOptions = list(epsi = NULL, typeSmallEM = NULL, typeEM = NULL, nberSmallEM = NULL, nberIterations = NULL, nberMaxIterations = NULL, putThreshold = NULL), cte = 1)

Arguments

xdata

A matrix of strings with the number of columns equal to ploidy * (number of variables).

K

The number of clusters (or populations).

S

The subset of clustering variables in the form of a vector of logicals indicating the selected variables. $S$ gathers variables that are not identically distributed in at least two clusters.

ploidy

The number of unordered observations represented by a string in xdata. For example, for genotypic data from diploid individual, $ploidy = 2$.

emOptions

A list of EM options (see EmOptions and setEmOptions).

cte

A double used as a value of $\lambda$ in the penalty function $pen(K,S)=\lambda*dim\left(K,S\right)$, where $dim\left(K,S\right)$ is the number of free parameters in the model defined by $\left(K,S\right)$.

Value

A list of

• N : The size (number of lines) of the dataset.
• K : The number of clusters (populations).
• S : A vector of logicals indicating the selected variables for clustering.
• dim : The number of free parameters.
• pi_K : The vector of mixing proportions.
• prob : A list of matrices, each matrix being the probabilities of a variable in different clusters.
• logLik : The log-likelihood.
• entropy : The entropy.
• criteria : Criteria values c(BIC, AIC, ICL, CteDim).
• Tik : A stochastic matrix given the a posteriori membership probabilities.
• mapClassif : Maximum a posteriori classification.
• NbersLevels : The numbers of observed levels of the considered categorical variables.
• levels : The observed levels.

References

• Dominique Bontemps and Wilson Toussile (2013) : Clustering and variable selection for categorical multivariate data. Electronic Journal of Statistics, Volume 7, 2344-2371, ISSN.
• Wilson Toussile and Elisabeth Gassiat (2009) : Variable selection in model-based clustering using multilocus genotype data. Adv Data Anal Classif, Vol 3, number 2, 109-134.

See Also

dataR2C for transformation of a classic data frame, backward.explorer, selectK.R, dimJump.R, model.selection.R for both model selection and classification.

Examples

data(genotype1)
head(genotype1)
genotype2 = cutEachCol(genotype1[, -11], ploidy = 2)
head(genotype2)

#See the EM options
EmOptions() # Options can be set by \code{\link{setEmOptions()}}
par5 = em.cluster.R (genotype2, K = 5, S = c(rep(TRUE, 8), rep(FALSE, 2)), ploidy = 2)
slotNames(par5)
head(par5["membershipProba"])
par5["mixingProportions"]
par5