README.md

In aczepielik/CrossTabCluster: Tools For Clsutering Cross Tables

CrossTabCluster

The package accompanies author's master thesis Clustering of Contingency tables. See r citation("CrossTabCluster") for details.

Installation

From source

The tar.gz archive of the source code of the package accompanies the thesis pdf file in the University archive. To install the package from source one can put the archive in the working directory of R session and prompt

install.packages('CrossTabCluster_0.1.0.tar.gz', repos = NULL, type = 'source')

in the R console.

From GitHub

The same version can be installed from author's GitHub with the remotes package:

remotes::install_github('aczepielik/CrossTabCluster@v0.1.0')

Usage

To load the package prompt:

library(CrossTabCluster)

Distances

Function chisq_distance returns a matrix of chi-square distances between row or column profiles of the contingency table. For example,

chisq_distance(ksarakil, dimension = 1)

calculates distances between row profiles of the attached ksarakil table.

To calculate distances between clusters as defined by Ward's method one can use function cluster_distance. For example

cluster_distance(
    israeli_survey,
    list(1, 2, c(3, 4), c(5, 6, 7), 8),
    dimension = 1
)

The same syntax is used to calculate the generalized distance between clusters:

generalized_distance(
    israeli_survey,
    list(1, 2, c(3, 4), c(5, 6, 7), 8),
    dimension = 1
)

Clustering

Two clustering algorithms described in this thesis, in a version suited for contingency tables, can be called with hclust_table and kmeans_table. Although their input syntax, function(table, dimension) differs from regular versions hclust and kmeans, the output is of the same class and can be processed as usual.

Evaluation of clustering solutions

Besides generalized distance between clusters CrossTabCluster gives opportunity to calculate silhouette scores (function(table, dimension, clusters)) and AIC (aic(table, dimension, clusters, scheme)).

aczepielik/CrossTabCluster documentation built on June 19, 2020, 7:53 p.m.