kmlcov (version 1.0.1)

kmlcov-package: Clustering longitudinal data using the likelihood as a metric of distance

Description

'kmlcov' Cluster longitudinal data using the likelihood as a metric of distance. The generalised linear model allow the user to introduce covariates with different level effects (2 levels).

Overview

To cluster longitudinal data, 'kmlcov' implement an ECM type algorithm which assign the trajectories to the cluster which maximise the likelihood. It is possible to introduce covariates via the generalised linear model with different level effects (2 levels) all spedified in one formula. The package implements the plot function to produce the diagrams at the condition of not having more than 2 different effects (although the program can deal with more than two effects) for e.g. time and treatment or time and sex q.v. the help of linkglmClust or kmlCov. To plot the main trajectories with more than two effects we recommand to use ggplot of the package ggplot2. To cluster longitudinal data, 2 functions have to be remembered glmClust and kmlCov, the first run the algorithm for clustering one time and the second run the same algorithm multiple times with different starting conditions. The method is greatly sensitive to the initial conditions, we therefore recommand to use kmlCov although it takes much more time.

Details

Package:

kmlcov

Type:

Package

Version:

1.0.1

Date:

2013-08-21

Authors@R:

c(person("Mamoun O", "Benghezal", role = c("aut", "cre"), email = "[e-mail]"), person("Christophe", "Genolini", role = c("ctb"), email = "[e-mail]"), )

License:

GPL-2

Depends:

methods

Collate:

'functions4glmClust.R' 'GlmCluster.R' 'glmclust-internal.R' 'glmClust.R' 'kmlcov-package.R' 'kmlCov.R'

Index:

glmClust                Clustering longitudinal data
kmlCov                  Clustering longitudinal data from different
                        starting conditions
which_best              Seek the best partitions

Converge-class Class '"Converge"' GlmCluster-class Class '"GlmCluster"' KmlCovList-class Class '"KmlCovList"' addIndic Create the new formula with the indicator covariates affect_rand Affect randomly the individuals to the clusters getNomCoef Get the name of the coefficients in the 'glm' object according to the current cluster log_lik Calculate the log-likelihood lowcyclo Measures of creatinine and time among cardiac transplant patients. majIndica Calculate an indicator vector observance Measures of obsevance. plot,GlmCluster-method Plot the main trajectories plot,KmlCovList-method Plot the main trajectories predict_clust Creates a character string expression to calculate the predicted values rwFormula Rewrite the formula with all the covariates seperateFormula Separate the covariates in a formula show,Converge-method Method for function 'show'

See Also

kmlCov glmClust which_best