Tree-Clustering Comparison

This function compares the within-group variation for groups formed by tree partitioning and unconstrained clustering. The results are plotted and returned invisibly.

Usage

trclcomp(x, method = "com", km = TRUE, mrt = TRUE)

Arguments

x

Rpart object with method "mrt" – see rpart

method

The clustering method for the unconstrained clustering

.

km

If TRUE a K-Means clustering is compared with the multivariate tree partitioning.

mrt

If TRUE an additional K-Means clustering with a starting configuration based on the multivariate tree partitioning is generated.

Details

The within-group variation for groups formed by multivariate tree partitioning and unconstrained clusterings are compared for all sizes of the hierarchy of tree partitions.

Value

Returns a list (invisibly) of the within-tree and within-cluster variation for all tree sizes.

References

De'ath G. (2002) Multivariate Regression Trees : A New Technique for Constrained Classification Analysis. Ecology 83(4):1103-1117.

Examples

data(spider)
fit <- mvpart(data.matrix(spider[,1:12])~herbs+reft+moss+sand+twigs+water,spider)

trclcomp(fit)
#> 1 2 3 

#> MRT error                :  1 0.481 0.336 
#> MRT.Cluster error        :  1 0.436 0.3 
#> Cluster error            :  1 0.436 0.341