Partition-class {longitudinalData}

~ Class: Partition ~

Description

An object of class Partition is a partition of a population into subgroups. The object also contains some information like the percentage of trajectories in each group or some qualities criterion.

Objects from the Class

Objects are mainly intend to be created by some clustering methods (like k-means, fuzzy k-means, mixture modeling, latent class analysis,...)

Slots

nbClusters

[numeric]: number of groups, between 1 and 26

clusters

[vector(factor)]: vector containing the groups of each individual. Groups are in upper-case letters.

percentEachCluster

[vector(numeric)]: percentage of trajectories contained in each group.

postProba

[matrix(numeric)]: assuming that in each clusters C and for each time T, variable follow a normal law (mean and standard deviation of the variable at time T restricted to clusters C), then it is possible to compute the postterior probabilities of each individual (that is the probabilities that an individual has to belong to each clusters). These probabilities are hold in postProba.

postProbaEachCluster

[vector(numeric)]: for each clusters C, mean of the post probabilities to belong to C of the individual that effectively belong to C. A high percent means that the individual that are in this cluter realy meant to be here.

criterionValues

[vector(numeric)]: Value of the quality criterions used to evaluate the quality of the Clustering. See qualityCriterion for details.

details

[vector(character)]: hold different optionnal informations like the algorithm (if any) used to find the partition, the convergence time, the imputation methods, the starting condition. Examples: details=c(algorithm="kmeans",convergenceTime="3").

validation rules

A class Partition object must follow some rules to be valid:

• Slots should be either all empty, or all non empty.

• nbClusters has to be lower or equal to 26.

• clusters is a factor in LETTERS[1:nbCluster].

Construction

Class Partition objects are mainly constructed by some clustering methods (like k-means, fuzzy k-means, mixture modeling, latent class analysis,...). Neverdeless, it is also possible to construct them from scratch using the fonction partition.

Get [

Object["nbClusters"]

[numeric]: Gets the number of clusters (the value of the slot nbClusters)

Object["clusters"]

[vector(factor)]: Gets the cluster of each individual (the value of the slot clusters)

Object["clustersAsInteger"]

[vector(integer)]: Gets the cluster of each individual and turn them into integer

Object["percentEachClusters"]

[vector(numeric)]: Get the percent of individual in each clusters (the value of the slot nbClusters)

Object["postProbaEachClusters"]

[vector(numeric)]: Get the post probabilities for each clusters.

Object["postProba"]

[matrix(numeric)]: Get the post probabilities for each individual and each clusters.

Object["criterionValues"]

[vector(numeric)]: gives the values of all the criterion values (the value of the slot criterionValues)

Object["details"]

[vector(character)]: Get the values of the slot details.

Object["XcriterionX"]

[numeric]: Get the value of the criterion XcriterionX. It can be one of Calinski.Harabatz, Krzysztof.Calinski, Genolini.Calinski, Ray.Turi, Davies.Bouldin, BIC, AIC, AICc or random.

Object["XspecialX"]

[character]: Get the value named XspecialX in the slot details (probably one of multiplicity, convergenceTime, imputationMethod or algorithm.)

Setteur [<-

Object["multiplicity"]<-value

[numeric]: In the slot details, sets the values names multiplicity to value.

Object["convergenceTime"]<-value

[numeric]: In the slot details, sets the values names convergenceTime to value.

The others slot can not be change after the object creation.

Author

Christophe Genolini
1. UMR U1027, INSERM, Université Paul Sabatier / Toulouse III / France
2. CeRSME, EA 2931, UFR STAPS, Université de Paris Ouest-Nanterre-La Défense / Nanterre / France

References

[1] C. Genolini and B. Falissard
"KmL: k-means for longitudinal data"
Computational Statistics, vol 25(2), pp 317-328, 2010

[2] C. Genolini and B. Falissard
"KmL: A package to cluster longitudinal data"
Computer Methods and Programs in Biomedicine, 104, pp e112-121, 2011

See Also

Overview: longitudinalData-package
Classes: LongData
Methods: partition

Examples

############
### Building Partition

### number
part <- partition(rep(c(1,2,1,3),time=3))

### LETTERS
part <- partition(rep(c("A","B","D"),time=4),details=c(convergenceTime="3",multiplicity="1"))

### Others don't work
try(partition(rep(c("A","Bb","C"),time=3)))

#############
### Setteur and Getteur

### '['
part["clusters"]
part["clustersAsInteger"]
part["nbClusters"]

### '[<-'
part["multiplicity"] <- 2
(part)