plot.SECdistr {sn}

Plotting methods for classes SECdistrUv and SECdistrMv

Description

Plotting methods for classes SECdistrUv and SECdistrMv

Usage

## S4 method for signature 'SECdistrUv'
plot(x, range, probs, main, npt = 251, ...)

## S4 method for signature 'SECdistrMv'
plot(x, range, probs, npt, landmarks = "auto",
   main, comp, compLabs, data = NULL, data.par = NULL, gap = 0.5, ...) 

Arguments

x

an object of class SECdistrUv or SECdistrMv.

range

in the univariate case, a vector of length 2 which defines the plotting range; in the multivariate case, a matrix with two rows where each column defines the plotting range of the corresponding component variable. If missing, a sensible choice is made.

probs

a vector of probability values. In the univariate case, the corresponding quantiles are plotted on the horizontal axis; it can be skipped by setting probs=NULL. In the multivariate case, each probability value corresponds to a contour level in each bivariate plot; at least one probability value is required. See ‘Details’ for further information. Default value: c(0.05, 0.25, 0.5, 0.75, 0.95) in the univariate case, c(0.25, 0.5, 0.75, 0.95) in the multivariate case.

npt

a numeric value or vector (in the univariate and in the multivariate case, respectively) to assign the number of evaluation points of the distribution, on an equally-spaced grid over the range defined above. Default value: 251 in the univariate case, a vector of 101's in the multivariate case.

landmarks

a character string which affects the placement of some landmark values in the multivariate case, that is, the origin, the mode and the mean (or its substitute pseudo-mean), which are all aligned. Possible values: "proper", "pseudo", "auto" (default), "". The option "" prevents plotting of the landmarks. With the other options, the landmarks are plotted, with some variation in the last one: "proper" plots the proper mean value, "pseudo" plots the pseudo-mean, useful when the proper mean does not exists, "auto" plots the proper mean if it exists, otherwise it switches automatically to the pseudo-mean. See dp2cp for more information on pseudo-CP parameters, including pseudo-mean.

main

a character string for main title; if missing, one is built from the available ingredients.

comp

a subset of the vector 1:d, if d denotes the dimensionality of the multivariate distribution.

compLabs

a vector of character strings or expressions used to denote the variables in the plot; if missing, slot(object,"compNames") is used.

data

an optional set of data of matching dimensionity of object to be superimposed to the plot. The default value data=NULL produces no effect. In the univariate case, data are plotted using rug at the top horizontal axis, unless if probs=NULL, in which case plotting is at the bottom axis. In the multivariate case, points are plotted in the form of a scatterplot or matrix of scatterplots; this can be regulated by argument data.par.

data.par

an optional list of graphical parameters used for plotting data in the multivariate case, when data is not NULL. Recognized parameters are: col, pch, cex. If missing, the analogous components of par() are used.

gap

a numeric value which regulates the gap between panels of a multivariate plot when d>2.

...

additional graphical parameters

Value

an invisible list. In the univariate case the list has three components: the input object representing the distribution and two numeric vectors with the coordinates of the plotted density values. In the multivariate case, the first element of the list is the input object representing the distribution and all subsequent list elements are lists with components of the panels comprising the matrix plot; the elements of these sub-lists are: the vectors of x and y coordinates, the names of the variables, the density values at the (x,y) points, a vector of the density levels of the curves appearing in each panel plot, with the corresponding approximate probability content as a vector attribute.

Details

For univariate density plots, probs are used to compute quantiles from the appropriate distribution, and these are superimposed to the plot of the density function, unless probs=NULL. In the multivariate case, each bivariate plot is constructed as a collection of contour curves, one curve for each probability level; consequently, probs cannot be missing or NULL. The level of the density contour lines are chosen so that each curve circumscribes a region with the quoted probability, to a good degree of approssimation; for additional information, see Azzalini and Capitanio (2014), specifically Complement 5.2 and p.179, and references therein.

Methods

signature(x = "SECdistrUv")

Plot an object x of class SECdistrUv.

signature(x = "SECdistrMv")

Plot an object x of class SECdistrMv.

Author(s)

Adelchi Azzalini

References

Azzalini, A. with the collaboration of Capitanio, A. (2014). The Skew-Normal and Related Families. Cambridge University Press, IMS Monographs series.

See Also

makeSECdistr, summary.SECdistr, dp2cp

Examples

# d=1
f1 <- makeSECdistr(dp=c(3,2,5), family="SC", name="Univariate Skew-Cauchy")
plot(f1)
plot(f1, range=c(-3,40), probs=NULL, col=4)
#    
# d=2
Omega2 <- matrix(c(3, -3, -3, 5), 2, 2)    
f2 <-  makeSECdistr(dp=list(c(10,30), Omega=Omega2, alpha=c(-3, 5)),  
                family="sn", name="SN-2d", compNames=c("x1","x2"))
plot(f2)    
x2 <- rmsn(100, dp=slot(f2,"dp"))
plot(f2, main="Distribution 'f2'", probs=c(0.5,0.9), cex.main=1.5, col=2,
  cex=0.8, compLabs=c(expression(x[1]), expression(log(z[2]-beta^{1/3}))), 
  data=x2, data.par=list(col=4, cex=0.6, pch=5))