qq10: QQ plots on -log10 scale for p-values

In tobyjohnson/gtx: Genetics ToolboX

Description

Plots a QQ plot on a -log10 scale, appropriate for P-values. Component functions help build up more complex QQ plots.

Usage

qq10(p, pmin = NULL, alpha = 0.01, ...)
qq10.new(pmin)
qq10.envelope(n, pmin, alpha = 0.01, col = "grey")
qq10.points(p, ...)

Arguments

p	A vector of observed p-values
pmin	The smallest observed or expected p-value to include in the plot
alpha	The alpha level for the envelope for the null distribution
n	The number of p-values for the envelope for the null distribution
col	The colour for shading the envelope for the null distribution
...	Additional arguments passed to points for plotting the p-values

Details

The plot is drawn in the current screen on the current device.

If the default pmin = NULL is used, qq10 sets pmin to the smallest observed or expected p-value. If there are observed p-values much smaller than expected, this may not produce a nice plot, and the user should specify a value for pmin and note that the y axis will be truncated.

The 100(1-alpha)% envelope for the null distribution is approximate, and is based on asymptotics for the distribution of the expected P-value, pointwise for each value of the observed P-value.

qq10.points.fast uses an experimental heuristic to thin points in the overlapping part of the QQ plot to speed up plotting. Because this involves subsetting the points to be plotted, additional graphical parameters must be subsetted to match and hence must be passed explicitly.

Value

Nothing should be returned.

Author(s)

Toby Johnson [e-mail]

Examples

## single command for simple QQ plot
qq10(sort(runif(100)), col = rainbow(100))
## example of building a plot to compare samples from two distributions
p1 <- runif(1000)
p2 <- rbeta(1000, shape1 = 0.6, shape2 = 1)
qq10.new(min(c(p1, p2, 1/2000)))
qq10.envelope(1000, min(c(p1, p2, 1/2000)))
qq10.points(p1, pch = 21, bg = "yellow")
qq10.points(p2, pch = 21, bg = "red")
legend("topleft", pch = rep(21, 2), pt.bg = c("red", "yellow"),
       legend = c(expression(beta(0.6,1)), expression(beta(1,1))))
box()

tobyjohnson/gtx documentation built on Aug. 30, 2019, 8:07 p.m.