[R] ecdf plots, lines, and y values

[Martin Maechler]

>>>>> "UweL" == Uwe Ligges <ligges at statistik.uni-dortmund.de>
>>>>>     on Thu, 03 Jun 2004 21:36:48 +0200 writes:

    UweL> Monica Palaseanu-Lovejoy wrote:
    >> Hi,
    >> 
    >> I have a question for the group, perhaps someone can help me 
    >> figure this out.  I've already looked in the help files and they were 
    >> no help to me.  
    >> 
    >> I have a vector of values and I am plotting an ecdf graph. 
    >> 
    >> 1. How can i draw a continuous line through the ecdf points? (lines 
    >> and type for the plot with an ecdf object does not work)

    UweL> Well, the help page ?plot.ecdf and ?plot.stepfun is pretty clear and 
    UweL> tells us to use verticals = TRUE.
    UweL> From the help page:

    UweL> F10 <- ecdf(rnorm(10))
    UweL> plot(F10, verticals = TRUE, do.p = FALSE)

Thank you Uwe,
but maybe Monica doesn't want to really plot the ecdf, but
rather a "continuized" {polyline} version of it ?

That's a bit less obvious to do and can be achieved by noting
that the (x,y) values are the (x,y) vectors in F10's
environment and that plot() or lines() on a list uses the list's
$x and $y components, e.g.,

 F10 <- ecdf(rnorm(10))
 plot(F10, verticals = TRUE, do.p = FALSE)
 lines(as.list(environment(F10)), col = "blue")

But even more nicely:

 with(environment(F10), lines(x,y,  col = "blue"))

This last trick can also be applied for the following question:

    >> 2. Supposing I have this line drawn. I can add a vertical line of a 
    >> known x value which intersects the graph. How can I determine the 
    >> y value of the graph point that is intersected by the abline? 

## Construct a "linear interpolator" using approxfun:
polyF10 <- with(environment(F10), approxfun(x,y))

## 'test' it graphically, using curve():
plot(F10, verticals = TRUE, do.p = FALSE)
curve(polyF10(x), add = TRUE, col="red")

## now use it, e.g. for two abscissa values simultaneously

 x0 <- c(0, 0.5)
 y0 <- polyF10(x0)

 x.min <- with(environment(F10), min(x))
## or rather simply
 x.min <- par("usr")[1]

 segments(x0,    0, x0,y0, lty=3)
 segments(x.min,y0, x0,y0, lty=3)


This seems a pretty neat example of making use of the function
objects returned from ecdf() and approxfun().

Regards,  Martin Maechler