Using Color in R

Key ideas

• Color Spaces
  • RGB Additive Color Space
  • RGB Color Names
• R Color Basics
  • Color Definition
  • Palette Types and RColorBrewer
  • Points, Axes, Text, Lines, Legends, Background
• Color Tips
  • Image Processing
  • Color Deficiency/Blindness
  • Why don't screen colors match printout?
  • Presentations

• RGB Red-Green-Blue

• HSV Hue-Saturation-Value

• HLS Hue-Lightness-Saturation

• LAB coordinates in CIE LAB space

• R Package: help(package=colorspace)

Why did you learn the wrong primary colors in elementary school?

0.0 to 1.0, decimal 0 to 255, hexadecimal 00 to FF

rgb(1,0,0)   # red

[1] "#FF0000"

rgb(0,1,0)   # green

[1] "#00FF00"

rgb(0,0,1)   # blue

[1] "#0000FF"

rgb(1,0,1)   # magenta = red   + blue

[1] "#FF00FF"

rgb(1,1,0)   # yellow  = red   + green

[1] "#FFFF00"

rgb(0,1,1)   # cyan    = green + blue

[1] "#00FFFF"

rgb(1,0,0)    # red

[1] "#FF0000"

rgb(255,0,0, maxColorValue=255)

[1] "#FF0000"

col2rgb(c("black", "red", "green", "blue", "yellow", "white"))

      [,1] [,2] [,3] [,4] [,5] [,6]
red      0  255    0    0  255  255
green    0    0  255    0  255  255
blue     0    0    0  255    0  255

Color names and RGB values also in file C:\Program Files\R\R-3.0.2\etc\rgb.txt

# colors() or colours()
length(colors())

[1] 657

head(colors())

[1] "white"         "aliceblue"     "antiquewhite"  "antiquewhite1" "antiquewhite2" "antiquewhite3"

tail(colours())

[1] "yellow"      "yellow1"     "yellow2"     "yellow3"     "yellow4"     "yellowgreen"

colors()[grep("red", colors())]   # colors with "red"

 [1] "darkred"         "indianred"       "indianred1"      "indianred2"      "indianred3"     
 [6] "indianred4"      "mediumvioletred" "orangered"       "orangered1"      "orangered2"     
[11] "orangered3"      "orangered4"      "palevioletred"   "palevioletred1"  "palevioletred2" 
[16] "palevioletred3"  "palevioletred4"  "red"             "red1"            "red2"           
[21] "red3"            "red4"            "violetred"       "violetred1"      "violetred2"     
[26] "violetred3"      "violetred4"     

head(colors()[grep("blue", colors())], 25)  # 25 of 66 "blue" colors

 [1] "aliceblue"      "blue"           "blue1"          "blue2"          "blue3"         
 [6] "blue4"          "blueviolet"     "cadetblue"      "cadetblue1"     "cadetblue2"    
[11] "cadetblue3"     "cadetblue4"     "cornflowerblue" "darkblue"       "darkslateblue" 
[16] "deepskyblue"    "deepskyblue1"   "deepskyblue2"   "deepskyblue3"   "deepskyblue4"  
[21] "dodgerblue"     "dodgerblue1"    "dodgerblue2"    "dodgerblue3"    "dodgerblue4"   

7-page Color Chart of colors()

https://github.com/EarlGlynn/colorchart/raw/master/ColorChart.pdf

Details showing how to create chart in R: Color Chart in R

3D colors() Example

From Example 6a in ?scatterplot3d help in library(scatterplot3d)

A color in R can be specified in three ways:

• color name

• hex constant #000000 through #FFFFFF (a 24-bit color)

• index into palette, which is vector of color names and/or hex constants

palette()   # show default palette

[1] "black"   "red"     "green3"  "blue"    "cyan"    "magenta" "yellow"  "gray"   

par(mar=c(1,1,1,1))  # Reduce margins
barplot(rep(1,8), yaxt="n", col=1:8)

To use default palette if not currently the defined palette:

barplot(rep(1,8), yaxt="n", col=palette("default"))

N <- 20
palette(rainbow(N))   # Define N-color rainbow palette
palette()             # Show colors in current palette

 [1] "red"     "#FF4D00" "#FF9900" "#FFE500" "#CCFF00" "#80FF00" "#33FF00" "#00FF19" "#00FF66"
[10] "#00FFB2" "cyan"    "#00B3FF" "#0066FF" "#001AFF" "#3300FF" "#7F00FF" "#CC00FF" "#FF00E6"
[19] "#FF0099" "#FF004D"

par(mar=c(1,1,1,1))   # Reduce margins
barplot(rep(1,N), yaxt="n", col=1:N)

Function spelled either grey() or gray()

N <- 16                           # N = Number of grey levels
0:(N-1) / (N-1)                   # vector with values from 0 to 1

 [1] 0.00000 0.06667 0.13333 0.20000 0.26667 0.33333 0.40000 0.46667 0.53333 0.60000 0.66667 0.73333
[13] 0.80000 0.86667 0.93333 1.00000

palette(grey(0:(N-1) / (N-1)))    # Define N-shades of grey (256 is max)
palette()                         # Show colors in current palette

 [1] "black"   "#111111" "#222222" "gray20"  "#444444" "#555555" "gray40"  "#777777" "#888888"
[10] "gray60"  "#AAAAAA" "#BBBBBB" "gray80"  "#DDDDDD" "#EEEEEE" "white"  

par(mar=c(1,1,1,1))                     # Reduce margins
barplot(rep(1,3*N), yaxt="n", col=1:N)  # Three cycles

Plotting 3*N bars with palette of N shades of grey results in color recycling. R recycles colors whenever there are more objects to plot than colors.

heat.colors()

par(mar=rep(0,4))    # No margins
barplot(rep(1,16), yaxt="n", col=heat.colors(16))

cm.colors() cyan-magenta

par(mar=rep(0,4))    # No margins
barplot(rep(1,16), yaxt="n", col=cm.colors(16))

topo.colors()

par(mar=rep(0,4))    # No margins
barplot(rep(1,16), yaxt="n", col=topo.colors(16))

terrain.colors()

par(mar=rep(0,4))    # No margins
barplot(rep(1,16), yaxt="n", col=terrain.colors(16))

Use colorRampPalette() to define function that creates palette.

z <- outer(1:20, 1:20, function(x,y) sin(sqrt(x*y)/3))
PaletteFunction <- colorRampPalette(c("red", "white", "blue"), space = "Lab")
filled.contour(z, col = PaletteFunction(20))

Qualitative Palette

• All colors have same perceptual weight/importance
• Typical application: bar plot

Sequential Palette

• For coding numerical information in a range
• Typical application: heat map

Diverging Palette

• Like sequential palette but middle neutral value between two extremes

Details:

• Choosing Color Palettes for Statistical Graphics
• Choosing Colour palettes. Part I: Introduction (ggplot2 examples)

library(RColorBrewer)
table(brewer.pal.info$category)

 div qual  seq 
   9    8   18 

brewer.pal.info[brewer.pal.info$category == "qual",]

        maxcolors category
Accent          8     qual
Dark2           8     qual
Paired         12     qual
Pastel1         9     qual
Pastel2         8     qual
Set1            9     qual
Set2            8     qual
Set3           12     qual

par(mar=c(0,4,0,0))  # Smaller margins than default
display.brewer.all(type="qual")

par(mar=c(0,4,0,0))  # Smaller margins than default
display.brewer.all(type="seq")

par(mar=c(0,4,0,0))  # Smaller margins than default
display.brewer.all(type="div")

Many R objects can take on different colors:

• Points
• Lines
• Axes
• Text
• Legends
• Background

palette()[1:5]   # First five elements of default palette

[1] "black"  "red"    "green3" "blue"   "cyan"  

x <- -2:2
y <- x^2
plot(x,y, col=1:5, pch=CIRCLE<-16, cex=2)

Equivalent plot statements:

assuming default palette

plot(x,y, col=1:5, pch=CIRCLE<-16, cex=2)

plot(x,y, col=c("black", "red", "green3", "blue", "cyan"), pch=CIRCLE, cex=2)

plot(x,y, col=c("#000000", "#FF0000", "green3", 4, 5), pch=CIRCLE, cex=2)

See code on next slide.

# Points, Title, Axis Labels
plot(0:10, 0:10, col=0:10, pch=CIRCLE<-16, axes=FALSE,
     main="Chart Title", col.main="blue", cex.main=2,
     xlab="X axis", ylab="Y axis", col.lab="blue", cex.lab=1.5)

# Margin Text
mtext("Bottom",  BOTTOM<-1, col="red")
mtext("Left",    LEFT<-2,   col="green")

# Axes
axis(BOTTOM, col="red",   col.axis="black", cex.axis=1.5)
axis(LEFT,   col="green", col.axis="green", at=2*0:5, labels=paste(20*0:5),
     las=AXIS_LABEL_HORIZONTAL<-1)

See code on next slide.

library(RColorBrewer)

Time <- 0:120  # minutes

Types  <- c("solid", "dotted", "dashed")
Colors <- brewer.pal(3, "Dark2")   # qualitative palette

Periods <- data.frame(Period1=cos(2*pi*Time/120),
                      Period2=cos(2*pi*Time/ 90),
                      Period3=cos(2*pi*Time/150))

matplot(Periods, type="l", lwd=3,
        main="Expression Vs. Time", cex.main=1.5,
        xlab="Time[min]", ylab="Expression",
        col=Colors, lty=Types)
legend("top", c("120 min period", "90 min period", "150 min period"),
       col=Colors, lty=Types, lwd=3)

par(bg="wheat", mar=c(4,4,0,0))  # background color
x <- -2:2
y <- x^2
plot(x,y, col=1:5, pch=CIRCLE<-16, cex=2)

• Useful when copying and pasting metafile to PowerPoint since default is "transparent"

• Bioconductor package EBImage is an image processing toolbox for R.
• Supports colormode of Grayscale or Color.
• channel function handles color space conversions.
• Color colormode images have red, green, blue and possiby alpha channels.

Segmentation of cells in Introduction to EBImage documentation.

• About 1 in 12 have some sort of color deficiency.
• ~8% of men and ~0.5% of women are affected according to Wikipedia.

Ishihara Color Blindess Tests from ishiharatest.blogspot.com

• What numbers do you see? 12 at the left. 29 or 70 at the right?
• How the Color Deficient Person Sees the World

library(dichromat)
par(mfcol=c(1,2), mar=rep(0,4))
N <- 20
pie(rep(1,N), col=heat.colors(N))
pie(rep(1,N), col=dichromat(heat.colors(N)))

dichromat function collapses red-green color distinctions to approximate the effect of the two common forms of red-green color blindness: protanopia and deuteranopia

See code on next slide.

library(dichromat)
par(mar=c(1,2,2,1))
layout(matrix(1:6,ncol=1))
image(1:10,1,matrix(1:10, ncol=1),
     col=colorschemes$BrowntoBlue.10,
     main="Brown to Blue (10)", axes=FALSE)
image(1:100,1,matrix(1:100 ,ncol=1),
     col=colorRampPalette(colorschemes$BrowntoBlue.10,space="Lab")(100),
     main="Brown to Blue Ramp", axes=FALSE)
...
image(1:12,1,matrix(1:12, ncol=1),col=colorschemes$Categorical.12,
     main="Categorical (12)", axes=FALSE)
image(1:12,1,matrix(1:12, ncol=1),
     col=dichromat(colorschemes$Categorical.12, "deutan"),
     main="Categorical (12) -- deuteranopia", axes=FALSE)
image(1:12,1,matrix(1:12,ncol=1),
     col=dichromat(colorschemes$Categorical.12, "protan"),
     main="Categorical (12) -- protanopia", axes=FALSE)

• Every device has a gamut of colors that it can display.
• Color gamuts between devices often do not match.
  • How should a color be represented if it cannot be displayed on a device?
• Color conversions may not be perfect, e.g., RGB additive colors must be converted to CMYK (Cyan-Magenta-Yellow-Black) subtractive colors for printing.
• Paper/ink differences

• Avoid unnecessary use of color.
• Use bright colors with small graphics to make them stand out.
• Be consistent in use of color.
• Do NOT use color as only attribute to show difference. E.g., consider color and line type.

Also see: Public Speaking: Cool Color Commentary

• Wheels, Pyramids, and Spinning Tops: The Scientific Approach to Color, Exhibit at Linda Hall Library through March 14

• Choosing Color Palettes for Statistical Graphics, Wirtschaftsuniversität Wien

• Color Chart in R

• Colour for Presentation Graphics, Proceedings of 3rd International Workshop on Distributed Statistical Computing

• Using Color in Information Display Graphics, NASA

• Using RColorBrewer to Colour Your Figures in R, R-bloggers

• Why Should Engineers and Scientists be Worried About Color?, IBM