Computes agglomerative hierarchical clustering of the dataset.
clustering(data, metric="euclidean", method="ward", nb)
Expression matrix, genes on rows and samples on columns
Character string specifying the metric to be used for calculating dissimilarities between the columns of the matrix. This must be one of 'euclidean', 'manhattan', 'pearson', 'pearsonabs', 'spearman', 'spearmanabs', 'jaccard', 'dice'
Character string defining the clustering method. This must be one of 'average', 'single', 'complete', 'ward'
The number of classes for kmeans and PAM clustering (kcentroids)
An object of class 'agnes' representing the clustering. See 'agnes.object' for details.
Available metrics are (written for two vectors x and y):
Usual square distance between the two vectors.
Absolute distance between the two vectors
Pearson correlation distance. (1 - r)/2
Absolute Pearson correlation distance. 1 - abs(r)
Spearman rank correlation distance. (1 - r)/2
Absolute Spearlan rnak correlation distance. 1 - abs(r)
Jaccard distance on 0-1 matrix
Dice distance on 0-1 matrix
Available agglomerative methods are :
The distance between two clusters is the average of the dissimilarities between the points in one cluster and the points in the other cluster.
we use the smallest dissimilarity between a point in the first cluster and a point in the second cluster (nearest neighbor method).
we use the largest dissimilarity between a point in the first cluster and a point in the second cluster
Ward's agglomerative method
The weighted distance from the agnes package
computes a divise clustering
Perform either kmeans clustering if the distance is euclidean or PAM clustering. The number of classes nb has to be done.
Kaufman, L. and Rousseeuw, P.J. (1990). Finding Groups in Data: An Introduction to Cluster Analysis. Wiley, New York.
# NOT RUN {
data(marty)
c<-clustering(marty, metric="pearson", method="ward")
clustering.plot(c, title="Hierarchical Clustering\nPearson-Ward")
# }
Run the code above in your browser using DataLab