chclust: Constrained hierarchical clustering

In nsj3/rioja: Analysis of Quaternary Science Data

Constrained hierarchical clustering

Description

Constrained hierarchical clustering.

Usage

chclust(d, method = "coniss")

## S3 method for class 'chclust'
plot(x, labels = NULL, hang = 0.1, axes = TRUE,
      xvar=1:(length(x$height)+1), xlim=NULL, ylim=NULL, 
      x.rev = FALSE, y.rev=FALSE, horiz=FALSE, ...)

bstick(n, ...)

## S3 method for class 'chclust'
bstick(n, ng=10, plot=TRUE, ...)

Arguments

d	a dissimilarity structure as produced, for example, by dist or vegdist.
method	the agglomeration method to be used. This should be (an unambiguous abbreviation of) either "coniss" or "conslink".
x, n	a constrained cluster object of class chclust produced by chclust.
xvar	numeric vector containing x-coordinates for the leaves of the dendrogram (see details below).
x.rev, y.rev	logical flags to reverse the x- or y-axis (and dendrogram labels). Defaults to FALSE.
horiz	logical indicating if the dendrogram should be drawn horizontally or not. Note that y-axis still refers to the dendrogram height even after rotating.
xlim, ylim	optional x- and y-limits of the plot, passed to the underlying plto function. The defaults for these show the full dendrogram.
labels, hang, axes	further arguments as in hclust.
ng	number of groups to display.
plot	logical to plot a broken stick model. Defaults to TRUE.
...	further graphical arguments. Use cex to change the text size of the x-axis labels, and cex.axis to change size of the y-axis values.

Details

chclust performs a constrained hierarchical clustering of a distance matrix, with clusters constrained by sample order. Returns an object of class chclust which can be plotted and interrogated. See Grimm (1987), Gordon & Birks (1972) and Birks & Gordon (1985) for discusssiom of the coniss and conslink algorithms. The resulting dendrogram can be plotted with plot. This is an extension of plot method for hclust that allows the dendrogram to be plotted horizontally or vertically (default). plot also accepts a numeric vector coordinates for x-axis positions of the leaves of the dendrogram. These could, for example, be the stratigraphic depths of core samples or geographic distances along a line transect.

bstick.chclust compares the dispersion of a hierarchical classification to that obtained from a broken stick model and displays the results graphically. See Bennett (1996) for details. bstick is a generic function and the default method is defined in package vegan. If package vegan is installed the function may be called using vegan::bstick, otherwise use bstick.chclust.

Value

Function chclust returns an object of class chclust, derived from hclust.

Author(s)

Steve Juggins

References

Bennett, K. (1996) Determination of the number of zones in a biostratigraphic sequence. New Phytologist, 132, 155-170.

Birks, H.J.B. & Gordon, A.D. (1985) Numerical Methods in Quaternary Pollen Analysis Academic Press, London.

Gordon, A.D. & Birks, H.J.B. (1972) Numerical methods in Quaternary palaeoecology I. Zonation of pollen diagrams. New Phytologist, 71, 961-979.

Grimm, E.C. (1987) CONISS: A FORTRAN 77 program for stratigraphically constrained cluster analysis by the method of incremental sum of squares. Computers & Geosciences, 13, 13-35.

See Also

hclust, cutree, dendrogram.

Examples

data(RLGH)
diss <- dist(sqrt(RLGH$spec/100))
clust <- chclust(diss)
bstick(clust, 10)
# Basic diagram
plot(clust, hang=-1)
# Rotated through 90 degrees
plot(clust, hang=-1, horiz=TRUE)
# Rotated and observations plotted according to sample depth.
plot(clust, xvar=RLGH$depths$Depth, hang=-1, horiz=TRUE, x.rev=TRUE)

# Conslink for comparison
clust <- chclust(diss, method = "conslink")
plot(clust, hang=-1)

nsj3/rioja documentation built on Sept. 17, 2024, 5:50 a.m.