partana: Partition Analysis

In optpart: Optimal Partitioning of Similarity Relations

Description

Partition analysis evaluates the within-cluster to among-cluster similarity of classifications as a measure of cluster validity

Usage

partana(c,dist)
## S3 method for class 'partana'
summary(object, ...)
## S3 method for class 'partana'
plot(x,panel='all',zlim=range(x$ptc),col=heat.colors(12),...)

Arguments

c	an integer or factor vector, or an object of class ‘clustering’, ‘partana’, ‘partition’, or ‘stride’
dist	an object of class ‘dist’ from functions dist, dsvdis or vegdist

.

object	an object of class ‘partana’
x	an object of class ‘partana’
panel	an integer switch to indicate which panel to draw
zlim	the min and max values for the color map
col	a color map name (heat.colors(12) is the default)
...	ancillary arguments to pass to summary or plot

Details

Calculates mean object-to-cluster similarity, mean cluster-to-cluster similarity, and mean within-cluster to among-cluster similarity. partana operates on partitions or clusterings produced by a wide range of algorithms, including specific methods for the products of functions optpart, slice, pam and diana.

summary produces a matrix of the mean cluster-to-cluster similarities, and the overall within-cluster/among-cluster similarity ratio.

plot plots two panels in sequence in the current device. The first shows the mean similarity of every object to each cluster, sorted by mean similarity to the other members of its own cluster, with objects as columns and clusters as rows. The second panel shows the mean similarity of every cluster to every other cluster and mean within-cluster similarity, ignoring cluster size. These plots are known as ‘Mondriaan’ plots, where the similarities are given by lines colored from min to max. If the ‘partana’ object was produced by optpart, a third panel is plotted showing the trace of the optimization.

Value

an object of class ‘partana’ with components:

ptc	matrix of mean object-to-cluster similarity
ctc	matrix of mean cluster-to-cluster similarity
clustering	vector of numeric cluster assignments
ratio	within-cluster to among-cluster similarity ratio

Author(s)

David W. Roberts [e-mail]

See Also

partition, optpart, plot.partana

Examples

data(shoshveg)
data(shoshsite)
dis.bc <- dsvdis(shoshveg,'bray/curtis')
demo.part <- partana(cut(shoshsite$elev,5),dis.bc)
summary(demo.part)

Example output

Loading required package: cluster
Loading required package: labdsv
Loading required package: mgcv
Loading required package: nlme
This is mgcv 1.8-33. For overview type 'help("mgcv-package")'.
This is labdsv 2.0-1
convert existing ordinations with as.dsvord()

Attaching package: ‘labdsv’

The following object is masked from ‘package:stats’:

    density

Loading required package: MASS
Loading required package: plotrix
Number of clusters =  5 

 1  2  3  4  5 
 5 34 44 36 31 

           [,1]       [,2]       [,3]       [,4]       [,5]
[1,] 0.16200952 0.11006768 0.08880058 0.03356128 0.02425765
[2,] 0.11006768 0.14977095 0.16528788 0.11541820 0.09821912
[3,] 0.08880058 0.16528788 0.20533098 0.17753652 0.15870803
[4,] 0.03356128 0.11541820 0.17753652 0.22051881 0.22725189
[5,] 0.02425765 0.09821912 0.15870803 0.22725189 0.28553848

Ratio of Within-cluster similarity/Among-cluster similarity =  1.4 

optpart documentation built on March 26, 2020, 6:18 p.m.