plot.lnre {zipfR}

Plot LNRE Population Distribution (zipfR)

Description

Visualisation of LNRE population distribution, showing either the (log-transformed) type or probability density function or the cumulative probability distribution function.

Usage

## S3 method for class 'lnre'
plot(x, y, ...,
     type=c("types", "probability", "cumulative"),
     xlim=c(1e-9, 1), ylim=NULL, steps=200,
     xlab=NULL, ylab=NULL, legend=NULL, grid=FALSE,
     main="LNRE Population Distribution",
     lty=NULL, lwd=NULL, col=NULL, bw=zipfR.par("bw"))

Arguments

x, y, ...

one or more objects of class lnre, containing trained LNRE models describing the population(s) to be plotted. Alternatively, all models can be passed as a list in the x argument if the method is called explicitly (see ‘Examples’).

type

what type of plot should be drawn, "types" for the log-transformed type density function, "probability" for the log-transformed probability density function, and "cumulative" for the cumulative probability distribution.

xlim, ylim

visible range on x- and y-axis. The default ylim is [0, 1] for type="cumulative" and automatically chosen to fit the selected density curves for type="density". Note that the x-axis is always logarithmic and xlim should be chosen accordingly.

steps

number of steps for drawing curves (increase for extra smoothness)

xlab, ylab

labels for the x-axis and y-axis (with suitable defaults depending on type)

legend

optional vector of character strings or expressions specifying labels for a legend box, which will be drawn in the upper right-hand or left-hand corner of the screen. If legend=TRUE, labels showing model type and parameters are automatically generated.

grid

whether to display a suitable grid in the background of the plot

main

a character string or expression specifying a main title for the plot

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 populations shown in the plot: the values are not automatically recycled.

bw

if TRUE, draw plot in B/W style (default is the global zipfR.par setting)

Details

There are two useful ways of visualising a LNRE population distribution, selected with the type argument:

types

A plot of the type density function g(\pi) over the type probability \pi on a log-transformed scale (so that the number of types corresponds to an integral over \log_{10} \pi, see ltdlnre). The log transformation is essential so that the density function remains in a reasonable range; a logarithmic y-axis would be very counter-intuitive. Note that density values correspond to the number of types per order of magnitude on the x-axis.

probability

A plot of the probability density function \pi g(\pi) over the type probability \pi on a log-transformed scale (so that probability mass corresponds to an integral over \log_{10} \pi, see ldlnre). Note that density values correspond to the total probability mass of types across one order of magnitude on the x-axis.

cumulative

A plot of the cumulative probability distribution, i.e. the distribution function F(\rho) = P(\pi \le \rho) showing the total probability mass of types with type probability \pi \le \rho. The x-axis shows \rho on a logarithmic scale (but is labelled more intuitively with \pi by default). No special transformations are required because 0 \le F(\rho) \le 1.

Line styles are defined globally through zipfR.par, but can be overridden with the optional parameters 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.

Other standard graphics parameters (such as cex or mar) cannot be passed to the plot function an need to be set up with par in advance.

See Also

lnre, ltdlnre, plnre zipfR.par, zipfR.plotutils

plot.tfl offers a different visualisation of the LNRE population distribution, in the form of a Zipf-Mandelbrot law rather than type density.

Examples

## visualise three LNRE models trained on same data
m1 <- lnre("zm", Dickens.spc)
m2 <- lnre("fzm", Dickens.spc)
m3 <- lnre("gigp", Dickens.spc)
plot(m1, m2, m3, type="types",
     xlim=c(1e-8, 1e-2), ylim=c(0, 7.5e4), legend=TRUE)
plot(m1, m2, m3, type="probability", 
     xlim=c(1e-8, 1e-2), grid=TRUE, legend=TRUE)

## cumulative probability distribution is not available for GIGP
plot(m1, m2, type="cumulative", grid=TRUE,
     xlim=c(1e-8, 1e-2), legend=c("ZM", "fZM"))

## first argument can also be a list of models with explicit call
models <- lapply(seq(.1, .9, .2), 
                 function (x) lnre("zm", alpha=x, B=.1))
plot.lnre(models, type="cum", grid=TRUE, legend=TRUE)
plot.lnre(models, type="prob", grid=TRUE, legend=TRUE)