R: Parallel-coordinate plot

parallel.plot {mining}

R Documentation

Parallel-coordinate plot

Description

Plots a data matrix using parallel coordinates.

Usage

parallel.plot(x,yscale=c("linear","range","none"),
              xscale=c("equal","linear","none"),proto=T,...)

Arguments

`x`	a data matrix or frame. Only numeric columns will be plotted.
`yscale`	describes how the columns should be scaled.
`xscale`	describes how the columns should be placed.
`proto`	If `TRUE` and the response variable is a factor, then only the mean value for each response is plotted. See `prototypes`.
`...`	Extra arguments to `labeled.curves`.

Details

Let the rows of the matrix be cases and the columns be dimensions. Each dimension is laid out as a vertical axis, and the value for each case is plotted as a point on that axis. Points corresponding to the same case are connected by lines. Thus each case is represented by a curve or profile, moving horizontally from dimension to dimension.

To enhance readability, the dimensions should usually be shifted to have a common center and scaled to have comparable units. This is controlled by yscale. If yscale="linear", the dimensions are automatically shifted and scaled by the linear profiles method. That is, the profiles are made to be as straight as possible. This may result in negative scaling for a dimension, i.e. the axis is reversed so that moving up means a lower value. When this happens, the dimension name is prepended with a minus sign. If yscale="range", the dimensions are shifted and scaled (by a positive number) so that the data ranges from 0 to 1 on each axis. If yscale="none", the dimensions are plotted as-is.

Another important choice is the ordering and spacing of the axes on the plot. If xscale="linear", the axes are placed according to the linear profiles method. If xscale="equal", the axes are placed similarly to "linear" but with the constraint that they must be equally spaced. If xscale="none", the axes are placed in the order that they appear in the matrix.

If the data frame has a categorical response, the profiles are colored according to the response. The color scheme can be adjusted using the arguments to labeled.curves.

Author(s)

Tom Minka

Examples

data(iris)
parallel.plot(iris)
parallel.plot(iris,proto=F,labels=NULL)

[Package Contents]