strucplot: Structured Displays of Contingency Tables

View source: R/strucplot.R

strucplotR Documentation

Structured Displays of Contingency Tables

Description

This modular function visualizes certain aspects of high-dimensional contingency tables in a hierarchical way.

Usage

strucplot(x, residuals = NULL, expected = NULL,
  condvars = NULL, shade = NULL, type = c("observed", "expected"),
  residuals_type = NULL, df = NULL, split_vertical = NULL, 
  spacing = spacing_equal, spacing_args = list(),
  gp = NULL, gp_args = list(),
  labeling = labeling_border, labeling_args = list(),
  core = struc_mosaic, core_args = list(),
  legend = NULL, legend_args = list(),
  main = NULL, sub = NULL, margins = unit(3, "lines"),
  title_margins = NULL, legend_width = NULL,
  main_gp = gpar(fontsize = 20), sub_gp = gpar(fontsize = 15),
  newpage = TRUE, pop = TRUE, return_grob = FALSE,
  keep_aspect_ratio = NULL, prefix = "", ...)

Arguments

x

a contingency table in array form, with optional category labels specified in the dimnames attribute.

residuals

optionally, an array of residuals of the same dimension as x (see details).

expected

optionally, an array of expected values of the same dimension as x, or alternatively the corresponding independence model specification as used by loglin or loglm (see details).

df

degrees of freedom passed to the shading functions used for inference. Will be calculated (and overwritten if specified) if both expected and residuals are NULL, or if expected is given a formula.

condvars

number of conditioning variables, if any; those are expected to be ordered first in the table. This information is used for computing the expected values, and is also passed to the spacing functions (see spacings).

shade

logical specifying whether gp should be used or not (see gp). If TRUE and expected is unspecified, a default model is fitted: if condvars is specified, a corresponding conditional independence model, and else the total independence model.

residuals_type

a character string indicating the type of residuals to be computed when none are supplied. If residuals is NULL, residuals_type must be one of "pearson" (default; giving components of Pearson's chi-squared), "deviance" (giving components of the likelihood ratio chi-squared), or "FT" for the Freeman-Tukey residuals. The value of this argument can be abbreviated. If residuals are specified, the value of residuals_type is just passed “as is” to the legend function.

type

a character string indicating whether the observed or the expected values of the table should be visualized.

split_vertical

vector of logicals of length k, where k is the number of margins of x (values are recycled as needed). A TRUE component indicates that the tile(s) of the corresponding dimension should be split vertically, FALSE means horizontal splits. Default is FALSE.

spacing

spacing object, spacing function, or a corresponding generating function (see details and spacings).

spacing_args

list of arguments for the spacing-generating function, if specified.

gp

object of class "gpar", shading function or a corresponding generating function (see details and shadings). Components of "gpar" objects are recycled as needed along the last splitting dimension. Ignored if shade = FALSE.

gp_args

list of arguments for the shading-generating function, if specified.

labeling

either a logical, or a labeling function, or a corresponding generating function (see details and labelings. If FALSE or NULL, no labeling is produced.

labeling_args

list of arguments for the labeling-generating function, if specified.

core

either a core function, or a corresponding generating function (see details). Currently, generating functions for mosaic plots (struc_mosaic), association plots (struc_assoc), and sieve plots (struc_sieve) are provided.

core_args

list of arguments for the core-generating function, if specified.

legend

either a legend-generating function, or a legend function (see details and legends), or a logical. If legend is NULL or TRUE and gp is a function, legend defaults to legend_resbased.

legend_args

list of arguments for the legend-generating function, if specified.

main

either a logical, or a character string used for plotting the main title. If main is a logical and TRUE, the name of the object supplied as x is used.

sub

a character string used for plotting the subtitle. If sub is a logical and TRUE and main is unspecified, the name of the object supplied as x is used.

margins

either an object of class "unit" of length 4, or a numeric vector of length 4. The elements are recycled as needed. The four components specify the top, right, bottom, and left margin of the plot, respectively. When a numeric vector is supplied, the numbers are interpreted as "lines" units. In addition, the unit or numeric vector may have named arguments (top, right, bottom, and left), in which case the non-named arguments specify the default values (recycled as needed), overloaded by the named arguments.

title_margins

either an object of class "unit" of length 2, or a numeric vector of length 2. The elements are recycled as needed. The two components specify the top and bottom title margin of the plot, respectively. The default for each specified title are 2 lines (and 0 else), except when a legend is plotted and keep_aspect_ratio is TRUE: in this case, the default values of both margins are set as to align the heights of legend and actual plot. When a numeric vector is supplied, the numbers are interpreted as "lines" units. In addition, the unit or numeric vector may have named arguments (top and bottom), in which case the non-named argument specify the default value (recycled as needed), overloaded by the named arguments.

legend_width

An object of class "unit" of length 1 specifying the width of the legend (if any). Default: 5 lines.

pop

logical indicating whether the generated viewport tree should be removed at the end of the drawing or not.

main_gp, sub_gp

object of class "gpar" containing the graphical parameters used for the main (sub) title, if specified.

newpage

logical indicating whether a new page should be created for the plot or not.

return_grob

logical. Should a snapshot of the display be returned as a grid grob?

keep_aspect_ratio

logical indicating whether the aspect ratio should be fixed or not. If unspecified, the default is TRUE for two-dimensional tables and FALSE otherwise.

prefix

optional character string used as a prefix for the generated viewport and grob names.

...

For convenience, list of arguments passed to the labeling-generating function used.

Details

This function—usually called by higher-level functions such as assoc and mosaic—generates conditioning plots of contingency tables. First, it sets up a set of viewports for main- and subtitles, legend, and the actual plot region. Then, residuals are computed as needed from observed and expected frequencies, where the expected frequencies are optionally computed for a specified independence model. Finally, the specified functions for spacing, gp, main plot, legend, and labeling are called to produce the plot. The function invisibly returns the "structable" object visualized.

Most elements of the plot, such as the core function, the spacing between the tiles, the shading of the tiles, the labeling, and the legend, are modularized in graphical appearance control (“grapcon”) functions and specified as parameters. For each element foo (= spacing, labeling, core, or legend), strucplot takes two arguments: foo and foo_args, which can be used to specify the parameters in the following alternative ways:

  1. Passing a suitable function to foo which subsequently will be called from strucplot to compute shadings, labelings, etc.

  2. Passing a corresponding generating function to foo, along with parameters passed to foo_args, that generates such a function. Generating functions must inherit from classes "grapcon_generator" and "}foo\code{".

  3. Except for the shading functions (shading_bar), passing foo(foo_args) to the foo argument.

  4. For shadings and spacings, passing the final parameter object itself; see the corresponding help pages for more details on the data structures.

If legends are drawn, a ‘cinemascope’-like layout is used for the plot to preserve the 1:1 aspect ratio.

If type = "expected", the expected values are passed to the observed argument of the core function, and the observed values to the expected argument.

Although the gp argument is typically used for shading, it can be used for arbitrary modifications of the tiles' graphics parameters (e.g., for highlighting particular cells, etc.).

Value

Invisibly, an object of class "structable" corresponding to the plot. If return_grob is TRUE, additionally, the plot as a grob object is returned in a grob attribute.

Note

The created viewports, as well as the tiles and bullets, are named and thus can conveniently be modified after a plot has been drawn (and pop = FALSE).

Author(s)

David Meyer David.Meyer@R-project.org

References

Meyer, D., Zeileis, A., and Hornik, K. (2006), The strucplot framework: Visualizing multi-way contingency tables with vcd. Journal of Statistical Software, 17(3), 1-48. doi: 10.18637/jss.v017.i03 and available as vignette("strucplot").

See Also

assoc, mosaic, sieve, struc_assoc, struc_sieve, struc_mosaic, structable, doubledecker, labelings, shadings, legends, spacings

Examples

data("Titanic")

strucplot(Titanic)
strucplot(Titanic, core = struc_assoc)
strucplot(Titanic, spacing = spacing_increase,
                   spacing_args = list(start = 0.5, rate = 1.5))
strucplot(Titanic, spacing = spacing_increase(start = 0.5, rate = 1.5))

## modify a tile's color
strucplot(Titanic, pop = FALSE)
grid.edit("rect:Class=1st,Sex=Male,Age=Adult,Survived=Yes",
          gp = gpar(fill = "red"))

vcd documentation built on June 9, 2022, 9:07 a.m.