eval.stability.clustering: Compares several clustering methods by means of its stability.

Description

This function compares several clustering methods (link functions and distances) by means of its stability.

Usage

eval.stability.clustering(X,nb=c(2:4),f=0.8,nsub=10,s0=0.98,
list_DIS=c("euclidean","pearson"),
list_ALGO=c("average","complete","ward"), pdfname = NULL,
verbose = TRUE)

Arguments

a data frame with p rows and n columns; if clustering on genes - samples by row and genes by column; if clustering on samples genes by row and samples by column

number of classes for partition; it must start at 2 and be sequential(by default 2,3 and 4)

part of the data set which is randomly picked for each subsample in the resampling procedure (by default 0.8)

nsub

half of the number of times the perturbation procedure is applied in the resampling procedure (by default 100)

list_DIS

the list of distances to test

list_ALGO

the list of linkage method to test

similarity threshold, must lie between 0 and 1 (by default 0.98)

pdfname

pdf file name for saving graphic, by default = NULL

verbose

print results if verbose = TRUE, by default = TRUE

Value

stab.methodsa list containing methods declared stable for each partition

Details

Resampling is done by randomly picking without replacement f of the data set; similarity threshold is the value which is pertinent to decide that two partitions are similar; see references

References

http://bioinfo-out.curie.fr/projects/cgh-clustering/index.html

Examples

Run this code

data(marty)
## Test on a smaller dataset
example.data<-marty[1:100,]
stab<-eval.stability.clustering(example.data)

Run the code above in your browser using DataLab