plot_tfl: Plot Type-Frequency List / Zipf Ranking (zipfR)

In zipfR: Statistical Models for Word Frequency Distributions

Description

Zipf ranking plot of a type-frequency list, or comparison of several Zipf rankings, on linear or logarithmic scale.

Usage

## S3 method for class 'tfl'
plot(x, y, ...,
     min.rank=1, max.rank=NULL, log="", 
     type=c("p", "l", "b", "o", "s"),
     xlim=NULL, ylim=NULL, freq=TRUE,
     xlab="rank", ylab="frequency", legend=NULL, grid=FALSE,
     main="Type-Frequency List (Zipf ranking)",
     bw=zipfR.par("bw"), cex=1, steps=200,
     pch=NULL, lty=NULL, lwd=NULL, col=NULL)

Arguments

x, y, ...	one or more objects of class tfl, containing the type frequency list(s) to be plotted. LNRE models of class lnre can also be specified to display the corresponding population Zipf rankings (see ‘Details’ for more information). It is also possible to pass all objects as a list in argument x, but the method needs to be called explicitly in this case (see ‘Examples’).
min.rank, max.rank	range of Zipf ranks to be plotted for each type-frequency list. By default, all ranks are shown.
log	a character string specifying the axis or axes for which logarithmic scale is to be used ("x", "y", or "xy"), similar to the log argument of plot.default.
type	what type of plot should be drawn. Types p (points), l (lines), b (both) and o (points over lines) are the same as in plot.default. Type s plots a step function with type frequencies corresponding to right upper corners (i.e. type S in plot); it is recommended for plotting full type-frequency lists and can be much more efficient than the other types. See ‘Details’ below.
xlim, ylim	visible range on x- and y-axis. The default values are automatically determined to fit the selected data in the plot.
freq	if freq=FALSE, plot relative frequency (per million words) instead of absolute frequency on the y-axis. This is useful for comparing type-frequency lists with different sample size and is required for plotting LNRE populations.
xlab, ylab	labels for the x-axis and y-axis.
legend	optional vector of character strings or expressions, specifying labels for a legend box, which will be drawn in the upper right-hand corner of the screen. If legend is given, its length must correspond to the number of type-frequeny lists in the plot.
grid	whether to display a suitable grid in the background of the plot (only for logarithmic axis)
main	a character string or expression specifying a main title for the plot
bw	if TRUE, draw plot in B/W style (default is the global zipfR.par setting)
cex	scaling factor for plot symbols (types "p", "b" and "o"). This scaling factor is not applied to other text elements in the plot; use par for this purpose.
steps	number of steps for drawing population Zipf rankings of LNRE models. These are always drawn as lines (regardless of type) and are not aligned with integer type ranks (because the LNRE models are actually continuous approximations).
pch, lty, lwd, col	style vectors that can be used to override the global styles defined by zipfR.par. If these vectors are specified, they must contain at least as many elements as the number of type-frequency lists shown in the plot: the values are not automatically recycled.

Details

The type-frequency lists are shown as Zipf plots, i.e. scatterplots of the Zipf-ranked frequencies on a linear or logarithmic scale. Only a sensible subset of the default plotting styles described in plot are supported: p (points), l (lines), b (both, with a margin around points), o (both overplotted) and s (stair steps, but actually of type S).

For plotting complete type-frequency lists from larger samples, type s is strongly recommended. It aggregates all types with the same frequency and is thus much more efficient than the other plot types. Note that the points shown by the other plot types coincide with the the right upper corners of the stair steps.

Trained LNRE models can also be included in the plot, but only with freq=FALSE. In this case, the corresponding population Zipf rankings are displayed as lines (i.e. always type l, regardless of the type parameter). The lines are intended to be smooth and are not aligned with integer type ranks in order to highlight the fact that LNRE models are continuous approximations of the discrete population.

Line and point styles are defined globally through zipfR.par, but can be overridden with the optional parameters pch, lty, lwd and col. In most cases, it is more advisable to change the global settings temporarily for a sequence of plots, though.

The bw parameter is used to switch between B/W and colour modes. It can also be set globally with zipfR.par.

See Also

tfl, vec2tfl, rlnre, spc2tfl, plot.spc, plot.vgc, plot.lnre, zipfR.par, zipfR.plotutils

Examples

## plot tiny type-frequency lists (N = 100) for illustration
tfl1 <- vec2tfl(EvertLuedeling2001$bar[1:100])
tfl2 <- vec2tfl(EvertLuedeling2001$lein[1:100])
plot(tfl1, type="b")
plot(tfl1, type="b", log="xy")
plot(tfl1, tfl2, legend=c("bar", "lein"))

## realistic type-frequency lists (type="s" recommended for efficiency)
tfl1 <- spc2tfl(BrownImag.spc)
tfl2 <- spc2tfl(BrownInform.spc)
plot(tfl1, tfl2, log="xy", type="s", 
     legend=c("fiction", "non-fiction"), grid=TRUE)
## always use freq=FALSE to compare samples of different size
plot(tfl1, tfl2, log="xy", type="s", freq=FALSE,
     legend=c("fiction", "non-fiction"), grid=TRUE)

## show Zipf-Mandelbrot law fitted to low end of frequency spectrum
m1 <- lnre("zm", BrownInform.spc)
m2 <- lnre("fzm", BrownInform.spc)
plot(tfl1, tfl2, m1, m2, log="xy", type="s", freq=FALSE, grid=TRUE,
     legend=c("fiction", "non-fiction", "ZM", "fZM"))

## call plot.tfl explicitly if only LNRE populations are displayed
plot.tfl(m1, m2, max.rank=1e5, freq=FALSE, log="xy")

## first argument can then also be a list of TFLs and/or LNRE models
plot.tfl(lapply(EvertLuedeling2001, vec2tfl), log="xy", type="s", freq=FALSE, 
         legend=names(EvertLuedeling2001))

Example output

zipfR documentation built on Jan. 8, 2021, 2:37 a.m.