PlotQQ: QQ-Plot for Any Distribution
In DescTools: Tools for Descriptive Statistics

PlotQQ

R Documentation

QQ-Plot for Any Distribution

Description

Create a QQ-plot for a variable of any distribution. The assumed underlying distribution can be defined as a function of f(p), including all required parameters. Confidence bands are provided by default.

Usage

PlotQQ(x, qdist=qnorm, main = NULL, xlab = NULL, ylab = NULL, datax = FALSE, add = FALSE,
       args.qqline = NULL, conf.level = 0.95, args.cband = NULL, ...)

Arguments

`x`	the data sample
`qdist`	the quantile function of the assumed distribution. Can either be given as simple function name or defined as own function using the required arguments. Default is `qnorm()`. See examples.
`main`	the main title for the plot. This will be "Q-Q-Plot" by default
`xlab`	the xlab for the plot
`ylab`	the ylab for the plot
`datax`	logical. Should data values be on the x-axis? Default is `FALSE`.
`add`	logical specifying if the points should be added to an already existing plot; defaults to `FALSE`.
`args.qqline`	arguments for the qqline. This will be estimated as a line through the 25% and 75% quantiles by default, which is the same procedure as `qqline()` does for normal distribution (instead of set it to `abline(a = 0, b = 1))`. The quantiles can however be overwritten by setting the argument `probs` to some user defined values. Also the method for calculating the quantiles can be defined (default is 7, see `quantile`). The line defaults are set to `col = par("fg")`, `lwd = par("lwd")` and `lty = par("lty")`. No line will be plotted if `args.qqline` is set to `NA`.
`conf.level`	confidence level for the confidence interval. Set this to `NA`, if no confidence band should be plotted. Default is `0.95`. The confidence intervals are calculated pointwise method based on a Kolmogorov-Smirnov distribution.
`args.cband`	list of arguments for the confidence band, such as color or border (see `DrawBand`).
`...`	the dots are passed to the plot function.

Details

The function generates a sequence of points between 0 and 1 and transforms those into quantiles by means of the defined assumed distribution.

Note

The code is inspired by the tip 10.22 "Creating other Quantile-Quantile plots" from R Cookbook and based on R-Core code from the function qqline. The calculation of confidence bands are rewritten based on an algorithm published in the package BoutrosLab.plotting.general.

Author(s)

Andri Signorell <andri@signorell.net>, Ying Wu <Ying.Wu@stevens.edu>

References

Teetor, P. (2011) R Cookbook. O'Reilly, pp. 254-255.

Examples

y <- rexp(100, 1/10)
PlotQQ(y, function(p) qexp(p, rate=1/10))

w <- rweibull(100, shape=2)
PlotQQ(w, qdist = function(p) qweibull(p, shape=4))

z <- rchisq(100, df=5)
PlotQQ(z, function(p) qchisq(p, df=5),
       args.qqline=list(col=2, probs=c(0.1, 0.6)),
       main=expression("Q-Q plot for" ~~ {chi^2}[nu == 3]))
abline(0,1)

# add 5 random sets
for(i in 1:5){
  z <- rchisq(100, df=5)
  PlotQQ(z, function(p) qchisq(p, df=5), add=TRUE, args.qqline = NA,
         col="grey", lty="dotted")
}

DescTools documentation built on Sept. 26, 2024, 1:07 a.m.