regr-package: Tools for Building Many Kinds of Regression Models

In regr0: Building regression models

Description

This package helps to perform efficient model building for many kinds of regression models, including (generalized) linear, robust linear, ordinal (aka polr) etc. It also contains some useful utility functions.

Details

Package:	regr
Type:	Package
Date:	>= 2008-08-20
License:	GPL
LazyLoad:	yes

The principal function regr() allows fitting regression models of many types by calling well-known R functions for

• Ordinary linear regression,

• logistic, Poisson, other generalized linear models, also for

• ordinal or nominal response variable.

The output of the function is somewhat different from the output of the base R functions lm, glm, polr etc. – it should be more informative for the task of model building.

The next most important function is the plot method for results of regr (see plot.regr), which gives more complete regression diagnostic plots than the specific plot methods of the mentioned “base” R functions.

An important part of residual analysis which is often neglected is plotting residuals versus explanatory variables, often called term plots. This is part of plot.regr, but can also be called independently by calling plresx.

If a regr result is printed, the output is more informative than the classical tables.

The package contains the methods for the usual extractor functions, like residuals, predict, ..., as well as for drop1, add1, etc, ...

The package contains a number of additional functions that should be useful for efficient data analysis. They include

• plresx() to plot residuals against explanatory variables,

• plres2x() to produce a plot showing interactions between two continuous regressors,

• plmatrix(), a version of pairs that displays any rectangular scatterplot matrix and can also be used to produce the full usual quadratic scatterplot matrix for many variables in appropriate junks with panels that are still large enough to be useful.

The package also includes some additional utility functions, like

• logst() for log transformation of variables with zeros,

• showd() for printing a "representative" part of a dataset.

For documenting the steps of the data analysis, there are several documenting functions:

tit and doc add attributes to data and results, which should allow for tracing the history of these objects (not yet thoroughly implemented).

stamp adds tracking information to plots, consisting of a project title, a step label, and the date when the plot is produced.

Some default options are added to options to reflect user preferences for output, see the help entries for stamp.

Author(s)

Werner A. Stahel, ETH Zurich, [e-mail]

References

More information is available at http://stat.ethz.ch/~stahel/regression

See Also

The car package.

Examples

  r.savings <- regr(sr ~ pop15 + pop75 + dpi + ddpi, data = LifeCycleSavings)
  plot(r.savings)

regr0 documentation built on May 2, 2019, 4:52 p.m.