00bertin-package: Bertin matrices

Description Details Note Author(s) References See Also Examples

Description

Package: bertin
Type: Package
Version: 0.1-xxx
SVN: $Revision: 74 $
First Release Date: 2007-12-08
This Release Date: 2011-09-09
License: LGPL

Bertin matrices are a means to analyze structures in a matrix using two-dimensional display techniques.

For a start, think of a Bertin matrix as a regular display, like in a spreadsheet. Each element of the data matrix is represented in a cell, giving a regular display grid. Instead of a numeric representation, the cell shows a symbolic image - a bar, or a typical token. Now feel fee to change the arrangement of rows or columns (i.e cases and variables, in conventional terms. Usually, the sequence of cases and variables is just accidental, and changes should be free. For the rare other cases, add an indicator to allow for recovery after permutation.

Bertin's strategy is to rearrange rows and columns to gain insight in the internal data structure,arranging variables and cases by similarity.

This package tries to provide some tools to analyze data using permutation techniques with the help of visualizations.

Details

The implementation is currently under reconstruction. The details are subject to change.

SVN version 0.1-72 or later: This is considered as an implementation model.

The basic functions try to allow as much of user control as possible.

An abstract data structure, currently matrices of class bertin, with attributes added to cache information that is presumably invariant over variations of display. The cache entries are bundled in two attributes:

Data cache (data_cache):

Main

Main identification

istransposed

For matrices: cases are columns (i.e. Bertin convention). Default is istransposed=FALSE, i.e. R conventions.

group

an optional factor, to be used as a grouping variable. If this is an integer, it is the index of the variable to be interpreted as group.

groupcolors

An optional colour palette used for group marks

scorerange

"by_var", "by_case", "global"

zscores

Scores. Always numeric, in R conventions, i.e. cases as rows.

Display cache (display_cache):

case_rowwise

Default: TRUE, i.e. R-conventions.

asp

Aspect Ratio

Data frames to be supported soon.

By convention, data are stored using R conventions, i.e. variables are in columns, cases in rows.

Before version 0.1-72, maybe unchanged:

At present, a Bertin matrix is a matrix of class bertin which has has attributes

scores

the matrix values. For example, ranks by row or by column. The original data matrix have to be transformed to comparable scales to allow for permutations. scores is a shadow matrix in transformed form.

rowmeans

a vector of mean values of scores, by row

colwmeans

a vector of mean values of scores, by columns

roworder

a permutation vector to define the arrangement of rows

colorder

a permutation vector to define the arrangement of columns

Note

We try to follow Bertin's original model, but consistency with R traditions suggest other choices occasionally.

Following Bertin, we display variables as rows as a default. This can be controlled by var.orientation.

We use a grey level palette as a default, whereas Bertin used black and white. Use palette=c("white","black") as an argument to get Bertin's choice.

Break points for colour choice are based on range. We use the mid-range as default cut point for a two colour choice, whereas Bertin uses the mean as default.

Author(s)

G. Sawitzki

Maintainer: G. Sawitzki <[email protected]>

References

J. Bertin: La graphique et le traitement graphique de l'information. Flammarion: Paris 1977.

A. de Falguerolles; F. Friedrich & G Sawitzki: A Tribute to J. Bertin's Graphical Data Analysis. Published in: W. Bandilla, F. Faulbaum (eds.): SoftStat '97. Advances in Statistical Software 6. Lucius & Lucius, Stuttgart 1997. (provided in the doc section of this package)

See Also

Package tableplot, package seriation.

Examples

1
2
3
4
5
6
data(Hotel)
image.bertin(bertinrank(Hotel), main= "Hotel data")

plot.bertin(bertinrank(Hotel), main= "Hotel data")

plot.bertin(Hotel, palette=c("white","black"))

bertin documentation built on May 31, 2017, 4:16 a.m.