In abarbour/cb.city: Access All of Cynthia Brewer's Color Palettes on cpt-city

library(cb.city)

Divergent color palettes should be used when the range of values is centered on some breakpoint -- zero, for example -- implying a difference or relative value. This would be appropriate for plotting model residuals, for example. The suite of color palettes currently available, by name, is

library(reshape2)
all.div <- list_cpt(pal.type=='div')
#http://www.edlazorvfx.com/ysu/html/ascii.html
shapes <- paste0("&#", c(10003, 9996, 9745, 9786, 9734, 9728), ";")
tbl <- dcast(all.div, pal.name ~ pal.n, 
             value.var='pal.n',
             fun.aggregate = function(x){lx <- length(x); as.character(ifelse(lx==1,sample(shapes,1),lx))}, margins = TRUE)
knitr::kable(tbl, "markdown")

where the top number gives the number of colors in the palette, and a happy character^[A happy character includes any of these: ✓, ✌, ☑, ☺, ☆, or ☀] indicates the palette (named to the far left) has that many colors. The columns on the far-right and bottom show counts of the number of TRUE variants of that given row or column.

What do these palettes look like?

Since there is some cost associated with downloading the palettes, here are just the length-11 palettes:

nt <- 11
cseq <- seq_len(nt)
shake <- FALSE
PLT <- function(pal.name, pal.n=nt, ...){
  if (shake){
    letters[1:pal.n]
  } else {
    cc. <- city_palette(pal.name, pal.n)
    plot(cc., ax=axis(1, at=cseq), ...)
    as.vector(cc.)    
  }
}
pals <- unique(all.div$pal.name)
layout(matrix(seq_len(length(pals))))
par(mar=c(0,4,0,0), oma=c(3,2,1,1), las=1)
Cols <- as.data.frame(t(sapply(pals, PLT)))
names(Cols) <- cseq

and a table of the hexidecimal codes:

knitr::kable(Cols)

Some examples

The red-to-blue (through white) is a classic palette that is highly effective. Use city_palette to quickly retrieve the color palette^[Remember: the result will be a cpt.cols object, so use as.vector or another method coercion to extract the color codes!]

cpal <- city_palette("RdBu",11)

or, with RColorBrewer:

cpal.rcb <- RColorBrewer::brewer.pal(11,"RdBu")
print(all.equal(as.vector(cpal), cpal.rcb))

From then on you can use the palette however you see fit. Here are two illustrations using that palette^[The plot.cpt.cols method is used to create the color scalebars.]: one is a pattern found by $\sin(x) + \cos(y)$, and the other a more interesting pattern based on the $\mathrm{sinc}$ function:

layout(matrix(c(1,2,1,2,1,2,1,2,3,4), nrow=5, byrow=TRUE))

par(mar=c(0,0.1,1,0.1), oma=c(0,0,0,0))

sines <- function(x, y){sin(x*pi/180)/2 + cos(y*pi/180)/2}
image(outer(0:180, 0:180, FUN = sines), col=as.vector(cpal), axes=FALSE, zlim=c(-1,1), asp=1)
mtext("A sum of sinusoids", side=3, line=0.)

sinc <- function(x,y){
  r. <- sqrt(x**2 + y**2) + .Machine$double.eps;
  sin(r.)/r.
}
x <- y <- seq(-10,10,by=0.1)
image(outer(x, y, FUN = sinc), col=as.vector(cpal), axes=FALSE, zlim=c(-1,1), asp=1)
mtext("the sinc function", side=3, line=0.)

par(mar=c(3,4,1,4))
plot(cpal, ax=axis(1, at=c(1,6,11), labels=c(-1,0,1)), ylab="")
plot(cpal, ax=axis(1, at=c(1,6,11), labels=c(-1,0,1)), ylab="")