Empirical Cumulative Distribution Function

Faster alternative for plotting the empirical cumulative distribution function (ecdf). The function offers the option to construct the ecdf on the base of a histogram, which makes sense, when x is large. So the plot process is much faster, without loosing much precision in the details.

PlotECDF(x, breaks = NULL, col = Pal()[1], ylab = "", 
         lwd = 2, xlab = NULL, ...)

Arguments

x: numeric vector of the observations for ecdf.
breaks: will be passed directly to hist. If left to NULL, no histogram will be used.
col: color of the line.
ylab: label for the y-axis.
lwd: line width.
xlab: label for the x-axis.
...: arguments to be passed to subsequent functions.

Details

The stats function plot.ecdf is fine for vectors that are not too large. However for n ~ 1e7 we would observe a dramatic performance breakdown (possibly in combination with the use of do.call).

PlotECDF is designed as alternative for quicker plotting the ecdf for larger vectors. If breaks are provided as argument, a histogram with that number of breaks will be calculated and the ecdf will use those frequencies instead of respecting every single point.
Note that a plot will rarely need more than ~1'000 points on x to have a sufficient resolution on usual terms. PlotFdist will also use this number of breaks by default.

Value

no value returned, use plot.ecdf if any results are required.

Author

Andri Signorell <andri@signorell.net>

Examples

PlotECDF(d.pizza$temperature)


# make large vector
x <- rnorm(n=1e7)

# plot only 1000 points instead of 1e7
PlotECDF(x, breaks=1000)