plot.adpc: Plotting a (low dimensional) ADPROCLUS solution

In adproclus: Additive Profile Clustering Algorithms

Plotting a (low dimensional) ADPROCLUS solution

Description

When passing a (low dimensional) ADPROCLUS solution of class adpc to the generic plot(), this method plots the solution in one of the following three ways:

Network

Each cluster is a vertex and the edge between two vertices represents the overlap between the corresponding clusters. The size of a vertex corresponds to the cluster size. The overlap is represented through color, width and numerical label of the edge. The numerical edge-labels can be relative (number of overlap observations / total observations) or absolute (number of observations in both clusters).

Profiles

Plot the profile matrix (P for full dimensional model, C for low dimensional model) in the style of a correlation plot to visualize the relation of each cluster with each variable.

Variables by components

Plot the low dimensional component-by-variable matrix B' in the style of a correlation plot to visualize the relation of each component with each original variable. NOTE: Only works for low dimensional ADPROCLUS.

Usage

## S3 method for class 'adpc'
plot(x, type = "Network", title = NULL, relative_overlap = TRUE, ...)

Arguments

x	Object of class adpc. (Low dimensional) ADPROCLUS solution
type	Choice for type of plot: one of "Network", "Profiles", "vars_by_comp". Default: "Network".
title	String. OPTIONAL.
relative_overlap	Logical, only applies to plot of type = "Network". If TRUE (default), the number of observations belonging to two clusters is divided by the total number of observations.
...	additional arguments will be passed on to the functions plot_cluster_network(), plot_profiles(), plot_vars_by_comp()

Value

Invisibly returns the input model.

Examples

# Loading a test dataset into the global environment
x <- stackloss

# Quick low dimensional clustering with K = 3 clusters and S = 1 dimensions
clust <- adproclus_low_dim(x, 3, 1)

# Produce three plots of the model
plot(clust, type = "Network")
plot(clust, type = "Profiles")
plot(clust, type = "vars_by_comp")

adproclus documentation built on Nov. 10, 2023, 1:07 a.m.