symbolmap: Graphics Symbol Map

Description

Create a graphics symbol map that associates data values with graphical symbols.

Usage

symbolmap(..., range = NULL, inputs = NULL,
                 transform = NULL, compress = transform, decompress = NULL)

Value

An object of class "symbolmap".

Arguments

...: Named arguments specifying the graphical parameters. See Details.
range: Optional. Range of numbers that are mapped. A numeric vector of length 2 giving the minimum and maximum values that will be mapped. Incompatible with inputs.
inputs: Optional. A vector containing all the data values that will be mapped to symbols. Incompatible with range.
transform: Optional. Experimental. An R function which should applied to input data for the symbol map before the graphics parameters ... are determined. See section on Nonlinear transformation.
compress: Optional. Experimental. An R function determining a nonlinear transformation of the domain of the symbol map, to be used when the symbol map is plotted. See section on Nonlinear transformation.
decompress: Experimental. An R function giving the inverse function of compress, if compress is specified.

Nonlinear transformation

The argument transform defines a transformation that will be applied to the input data for the symbol map.

If transform is given, it should be an R function. All input data for the symbol map will first be transformed by this function, before the graphical parameters specified by ... are determined.

A typical example would be a logarithmic symbol map defined by transform = log10 together with something like size = function(x) { 3 * x }. This would mean that a numerical value z will be represented on the plot by a circle of diameter size(transform(z)) = 3 * log10(z) on the physical scale of the plot.
The arguments compress and decompress define a transformation of the range of numbers that is used only when the symbol map is plotted.

A typical example would be a logarithmic symbol map defined by compress = log10 and decompress = function(x) { 10^x }.

The arguments compress and decompress have no effect on the interpretation of the other arguments. They only affect the way in which the symbol map is plotted by plot.symbolmap. For a continuous symbol map, the range of input values is plotted on the compressed scale, but annotated on the original scale. See the Examples.

If transform is given, then the default value of compress is the same function transform. This reflects the fact that, when the user has specified that the input data should be transformed to another scale in order to determine their graphical representation, it would usually be appropriate to display the symbol map legend on the same transformed scale. However this can be overridden by specifying another value for compress, including NULL.

The arguments transform, compress and decompress should be functions which are vectorised (i.e. if x is a vector then compress(x) and decompress(x) are also vectors of the same length as x) and increasing (if x < y then compress(x) < compress(y) and decompress(x) < decompress(y).

The argument decompress is not needed in the following cases:

If compress is the function log10, then decompress is taken to be its inverse function(x) { 10^x }.
If compress is a cumulative distribution function (of class "ecdf", "ewcdf" or "interpolatedCDF") then decompress is taken to be its inverse function decompress = quantilefun(compress).

Author

Adrian Baddeley Adrian.Baddeley@curtin.edu.au, Rolf Turner rolfturner@posteo.net and Ege Rubak rubak@math.aau.dk.

Details

A graphical symbol map is an association between data values and graphical symbols. The command symbolmap creates an object of class "symbolmap" that represents a graphical symbol map.

Once a symbol map has been created, it can be applied to any suitable data to generate a plot of those data. This makes it easy to ensure that the same symbol map is used in two different plots. The symbol map can be plotted as a legend to the plots, and can also be plotted in its own right.

The possible values of data that will be mapped are specified by range or inputs.

if range is given, it should be a numeric vector of length 2 giving the minimum and maximum values of the range of numbers that will be mapped. These limits must be finite.
if inputs is given, it should be a vector of any atomic type (e.g. numeric, character, logical, factor). This vector contains all the possible data values that will be mapped.
If neither range nor inputs is given, it is assumed that the possible values are real numbers.

The association of data values with graphical symbols is specified by the other arguments ... which are given in name=value form. These arguments specify the kinds of symbols that will be used, the sizes of the symbols, and graphics parameters for drawing the symbols.

Each graphics parameter can be either a single value, for example shape="circles", or a function(x) which determines the value of the graphics parameter as a function of the data x, for example shape=function(x) ifelse(x > 0, "circles", "squares"). Colourmaps (see colourmap) are also acceptable because they are functions.

Currently recognised graphics parameters, and their allowed values, are:

shape: The shape of the symbol: currently either "circles", "squares", "arrows", "crossticks" or NA. This parameter takes precedence over pch. (Crossticks are used only for point patterns on a linear network).
size: The size of the symbol: a positive number or zero.
pch: Graphics character code: a positive integer, or a single character. See par.
cex: Graphics character expansion factor.
cols: Colour of plotting characters.
fg,bg: Colour of foreground (or symbol border) and background (or symbol interior).
col,lwd,lty: Colour, width and style of lines.
etch: Logical. If TRUE, each symbol is surrounded by a border drawn in the opposite colour, which improves its visibility against the background. Default is FALSE.
direction,headlength,headangle,arrowtype: Numeric parameters of arrow symbols, applicable when shape="arrows". Here direction is the direction of the arrow in degrees anticlockwise from the \(x\) axis; headlength is the length of the head of the arrow in coordinate units; headangle is the angle subtended by the point of the arrow; and arrowtype is an integer code specifying which ends of the shaft have arrowheads attached (0 means no arrowheads, 1 is an arrowhead at the start of the shaft, 2 is an arrowhead at the end of the shaft, and 3 is arrowheads at both ends).

A vector of colour values is also acceptable for the arguments col,cols,fg,bg if range is specified.

Examples

Run this code

  g <- symbolmap(inputs=letters[1:10], pch=11:20)

  g1 <- symbolmap(range=c(0,100), size=function(x) x/50)

  ## squares and circles with area proportional to |x|
  ## For 'squares', size is side length; size = sqrt(area)
  ## For 'circles', size is diameter;    size = sqrt(area * 4/pi)
  g2 <- symbolmap(shape=function(x) ifelse(x > 0, "circles", "squares"),
                  size=function(x) sqrt(ifelse(x > 0,
                                               abs(x)*4/pi,
                                               abs(x))),
                  bg = function(x) ifelse(abs(x) < 1, "red", "black"))

  colmap <- colourmap(topo.colors(20), range=c(0,10))
  g3 <- symbolmap(pch=21, bg=colmap, range=c(0,10))
  plot(g3)

  ## logarithmic symbol map
  gm <- symbolmap(range=c(1,1000), pch=21, transform=log10,
                  size=function(x) { x + 1 })
  ## input value x is plotted as a circle of diameter = log10(x) + 1.
  gm(100)
  plot(gm, nsymbols=4)
  ## note logarithmic scale of legend, because compress=transform by default.
  
  ## logarithmic display scale only
  gl <- symbolmap(range=c(1,1000), pch=21, compress=log10)
  gl(10)
  plot(gl, nsymbols=4)
  gu <- symbolmap(range=c(1,1000), pch=21)
  gu(10)
  plot(gu, nsymbols=4)
  ## log transformation 'compress' does not affect symbol map itself

Run the code above in your browser using DataLab