# 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).

### 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 = "mobenghezal@gmail.com"), person("Christophe", "Genolini", role = c("ctb"), email = "christophe.genolini@u-paris10.fr"), ) |

License: | GPL-2 |

Depends: | methods |

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

Index:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 | ```
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'
``` |

### 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.

### See Also

kmlCov

glmClust

which_best