Functions and Methods for Correspondence Regression

Share:

Description

This package provides functions and methods for performing correspondence regression, i.e. the correspondence analysis of the crosstabulation of a categorical variable Y in function of another one X, where X can in turn be made up of the combination of various categorical variables. Consequently, correspondence regression can be used to analyze the effects for a polytomous or multinomial outcome variable. The central function in the package is corregp, which enables methods for printing, summarizing and plotting the output. Additionally, there are functions for computing confidence intervals, ellipses or 3D ellipsoids (by means of bootstrapping).

Contents

This package consists of the following datasets, functions, generics and methods:

Datasets

  • HairEye Hair and eye color of statistics students (data frame).

  • COMURE The use of linguistic variants in translations vs. non-translations and in six different registers.

  • TSS The use of inflected or uninflected determiners in vernacular Belgian Dutch.

Functions

  • ci A helper function to compute confidence intervals on the basis of a numeric vector.

  • confGet An internal function to retrieve the conf components in a corregp object.

  • corregp The basic function to perform correspondence regression. Typically, one starts here, and then one uses print, summary, screeplot or plot methods.

Generics

  • cint Compute confidence intervals.

  • cell Compute confidence ellipses.

  • cell3d Compute 3D confidence ellipsoids.

  • ciplot Plot confidence intervals.

  • pcplot Plot parallel coordinates.

Methods

  • print.corregp Print the output of a correspondence regression.

  • summary.corregp Give a summary of a correspondence regression.

  • print.summary.corregp Print the summary of a correspondence regression.

  • screeplot.corregp Make a scree plot on the basis of the output of a correspondence regression.

  • cint.corregp Compute confidence intervals on the basis of the output of a correspondence regression. Typically, this function is not so much used directly as it is called by a ciplot.corregp command.

  • ciplot.corregp Plot confidence intervals on the basis of the output of a correspondence regression.

  • pcplot.corregp Make a parallel coordinate plot on the basis of the output of a correspondence regression.

  • cell.corregp Compute confidence ellipses on the basis of the output of a correspondence regression. Typically, this function is not so much used directly as it is called by a plot.corregp command.

  • plot.corregp Plot the output (and the optional confidence ellipses) of a correspondence regression.

  • cell3d.corregp Compute 3D confidence ellipsoids on the basis of a correspondence regression. Typically, this function is not so much used directly as it is called by a plot3d.corregp command.

  • plot3d.corregp Plot the 3D output (and the optional confidence ellipsoids) of a correspondence regression.

Future prospects

  • Include a vignette.

  • Allow for more than one random factor (e.g. enable the argument parm to be a character vector of more than one element).

  • Specify methods for anova, predict, residuals etc.

  • Enable scale transformations for all plots (and corresponding confidence regions).

  • Provide the possibility for so-called "calibration lines".

Author

Koen Plevoets, koen.plevoets@ugent.be

Acknowledgements

This package has benefited greatly from the helpful comments of Isabelle Delaere and Gert De Sutter. Thanks to Kurt Hornik and Uwe Ligges for proofing this package.