overlaid filled contour plots: solved
MerryXmas() to R-help!
No one answered this query, but I found an easy solution that others may
find useful, using a wrapper for contour() that calls contourLines()
and then replots the contours using polygon(). It could use a little
tuning (easier way to specify fill.alpha) to make it a general extension
of contour() or perhaps incorporate the additional fill.* arguments
directly into contour, where they probably belong.
# Add facility to fill contours in contour() and to return
# the contourLines()
# All arguments of contour.default commented out can be
# passed to contour() as ...
# fill.col is a vector of fill colors corresponding to levels.
# if not supplied, calculate a set of transparent colors
# based on col and fill.alpha
# TODO: add logic for fill.col=NULL to nothing extra, other
# than return contourLines
# TODO: make fill.alpha accept a more standard 0-1 numeric
contourf <- function(
x = seq(0, 1, length.out = nrow(z)),
y = seq(0, 1, length.out = ncol(z)),
z,
nlevels = 10,
levels = pretty(zlim, nlevels),
# labels = NULL,
# xlim = range(x, finite = TRUE),
# ylim = range(y, finite = TRUE),
zlim = range(z, finite = TRUE),
# labcex = 0.6, drawlabels = TRUE, method = "flattest",
# vfont, axes = TRUE, frame.plot = axes,
col = par("fg"),
fill.col,
fill.alpha = "80", # hex value for alpha transparency
# lty = par("lty"), lwd = par("lwd"),
add = FALSE, ...) {
contour(x,y,z, nlevels=nlevels, levels=levels, zlim=zlim, col=col,
add=add, ...)
line.list <- contourLines(x, y, z, nlevels=nlevels, levels=levels)
# contourLines returns a list of lists, each with components
# 'level', 'x', 'y'
if(missing(fill.col)) {
colramp = colorRampPalette(c("white", col))
fill.col <- paste(colramp(nlevels+1), fill.alpha, sep="")
}
for (i in seq_along(line.list)) {
polygon(line.list[[i]][2:3], col=fill.col[i], border=NA)
}
invisible(line.list)
}
best,
-Michael
On 12/22/2011 9:54 AM, Michael Friendly wrote:
I'm trying to make a set of contour plots of bivariate kernel density estimates, showing three such plots overlaid, similar to this plot http://euclid.psych.yorku.ca/SCS/Private/Test/ridge-boot2.pdf except that I would like to have the contours *filled* (using transparent colors). To make this reproducible, I've saved the results of KernSmooth::bkde2D() in the following file: http://euclid.psych.yorku.ca/SCS/Private/Test/bkde2D-boot.RData From this, the essence of my plot can be produced by the following: # show calls to bkde2D() #library(KernSmooth) #dest0 <- bkde2D(lbt[,wh], bandwidth=.5, range.x=list(c(-16, 8), c(-5,5))) #dest1 <- bkde2D(lbt[,1+wh], bandwidth=.5, range.x=list(c(-16, 8), c(-5,5))) #dest2 <- bkde2D(lbt[,2+wh], bandwidth=.5, range.x=list(c(-16, 8), c(-5,5))) #save(dest0, dest1, dest2, cm0, cm1, cm2, file="bkde2D-boot.RData") con <- url("http://euclid.psych.yorku.ca/SCS/Private/Test/bkde2D-boot.RData") ## print the value to see what objects were saved. print(load(con)) close(con) col <- c("black", rainbow(5, start=.6, end=.1)) clr <- col[c(1,3,6)] contour(dest0$x1, dest0$x2, dest0$fhat, nlevels=4, xlim=c(-8,4), ylim=c(-3,0), cex.lab=1.25, xlab="GNP", ylab="Unemployed" ) contour(dest1$x1, dest1$x2, dest1$fhat, nlevels=4, add=TRUE, col=clr[2]) contour(dest2$x1, dest2$x2, dest2$fhat, nlevels=4, add=TRUE, col=clr[3]) lines(rbind(cm0, cm1, cm2)) points(rbind(cm0, cm1, cm2), col=clr, pch=16, cex=1.5) I've tried filled.contour(), as below, except that (a) it doesn't allow add=TRUE, and (b) has no way to suppress the legend. # try filled contours, using transparent colors colramp0 = colorRampPalette(c("white", clr[1])) col <- paste(colramp0(5), "80", sep="") # make transparent filled.contour(dest0$x1, dest0$x2, dest0$fhat, nlevels=4, xlim=c(-8,4), ylim=c(-3,0), cex.lab=1.25, col=col, xlab="GNP", ylab="Unemployed" ) colramp1 = colorRampPalette(c("white", clr[2])) col <- paste(colramp1(5), "80", sep="") # make transparent filled.contour(dest1$x1, dest1$x2, dest1$fhat, nlevels=4, xlim=c(-8,4), ylim=c(-3,0), cex.lab=1.25, col=col, xlab="GNP", ylab="Unemployed" ) colramp2 = colorRampPalette(c("white", clr[3])) col <- paste(colramp2(5), "80", sep="") # make transparent filled.contour(dest2$x1, dest2$x2, dest2$fhat, nlevels=4, xlim=c(-8,4), ylim=c(-3,0), cex.lab=1.25, col=col, xlab="GNP", ylab="Unemployed" ) Note that the structure of the result of bkde2D() is a list of the following form, so I think I'm stuck with plotting the 3 density estimates separately as I do above.
> str(dest0)
List of 3 $ x1 : num [1:51] -16 -15.5 -15 -14.6 -14.1 ... $ x2 : num [1:51] -5 -4.8 -4.6 -4.4 -4.2 -4 -3.8 -3.6 -3.4 -3.2 ... $ fhat: num [1:51, 1:51] 2.08e-04 1.60e-04 5.88e-05 1.08e-05 1.20e-06 ...
>
Any ideas? -Michael