That code has some problems if your time series do not cross at the time
points. E.g., Duncan's code with some diagnostic plotting is:
f0 <- function (site1, site2, x = time(site1), check = FALSE)
{
stopifnot(length(x) == length(site1), length(x) == length(site2), all(diff(x) > 0))
if (check) {
oldMfrow <- par(mfrow = c(2, 1))
on.exit(par(oldMfrow))
matplot(x, cbind(site1, site2), type = "b")
}
smaller <- pmin(site1, site2)
plot(x, site1, ylim = range(c(site1, site2)), type = "n")
polygon(c(x, rev(x)), c(smaller, rev(site1)), col = "red")
polygon(c(x, rev(x)), c(smaller, rev(site2)), col = "blue")
}
# See how it messes the crossings in
f0(c(0, 1, -1, 0), c(1, 0, -2, 1), check=TRUE)
The following is a quick and dirty way of adding all the crossing
points to the original data sets so the plot is right.
addCrossingPoints <- function (y1, y2, x)
{
stopifnot(length(y1)==length(x), length(y2)==length(x), all(diff(x)>0))
# isCrossingSegment[i]==TRUE means that y1 and y2 cross between
# x[i] and x[i+1] (strictly, not at x[i] or x[i+1]).
i <- seq_len(length(y1) - 1)
isCrossingSegment <- ((y1[i] < y2[i]) & (y1[i+1] > y2[i+1])) |
((y1[i] > y2[i]) & (y1[i+1] < y2[i+1]))
if (any(isCrossingSegment)) {
i <- which(isCrossingSegment)
dx <- x[i+1] - x[i]
m1 <- (y1[i+1] - y1[i]) / dx
m2 <- (y2[i+1] - y2[i]) / dx
newDX <- (y1[i] - y2[i]) / (m2 - m1)
newY <- y1[i] + newDX * m1
o <- order( x <- c(x, x[i] + newDX))
x <- x[o]
y1 <- c(y1, newY)[o]
y2 <- c(y2, newY)[o]
}
list(y1=y1, y2=y2, x=x)
}
# Combine it with Duncan's code to get a better looking plot
f1 <- function(site1, site2, x=time(site1), check = FALSE) {
tmp <- addCrossingPoints(site1, site2, x)
f0(tmp$y1, tmp$y2, tmp$x, check=check)
}
f1(c(0, 1, -1, 0), c(1, 0, -2, 1), check=TRUE)
Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com
-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf
Of Roland Pape
Sent: Friday, June 07, 2013 2:51 AM
To: Duncan Murdoch
Cc: r-help at r-project.org
Subject: Re: [R] Conditional coloring of area between curves
Hi Duncan,
that's exactly what I was looking for! The series have common times, so that's no
problem. Thanks a lot!
Roland
________________________________________
Von: Duncan Murdoch [murdoch.duncan at gmail.com]
Gesendet: Donnerstag, 6. Juni 2013 18:45
An: Roland Pape
Cc: r-help at r-project.org
Betreff: Re: [R] Conditional coloring of area between curves
On 06/06/2013 10:41 AM, Roland Pape wrote:
Dear list,
I have two time series of temperatures from different sites plotted in the same diagram
and would like to color the area between the curves differently, dependent on whether
site 1 is cooler than site 2 (colored let's say in blue) or warmer (red). While polygone()
works fine to color the enclosed area in general, I'm struggling with this conditional
coloring. Any help is greatly appreciated!
Suppose the series are named "site1" and "site2", and they have common
times in a variable "times". Then the following should do what you want:
smaller <- pmin(site1, site2)
plot(x, site1, ylim = range(c(site1, site2)), type="n")
polygon(c(x, rev(x)), c(smaller, rev(site1)), col="red")
polygon(c(x, rev(x)), c(smaller, rev(site2)), col="blue")
If the times for the two series are different it's a little harder;
first you need to give them common times, and that will depend on how
you decide to evaluate the values between observations. Probably linear
interpolation (using approx()) is fine, but it's up to you.
Duncan Murdoch
______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.