Overview

We will cover how to “(c) manually set some scales” for a plot, focusing on

• Position scales
• Colour scales
• Manual scales

The contents for (c) are based on Chapters 5 and 6 of the book “ggplot2: elegant graphs for data analysis” by Hadley Wickham.

Overview

Scales

• control mapping from data to aesthetics (e.g., size, colour, position, or shape)
• provide guides (e.g., axes and legends)
• can be roughly divided into four categories: position scales, colour scales, the manual discrete scale, and the identity scale; for position aesthetics, axes are guides; for all other aesthetics, legends do the job

A scale is needed for each plot, and ggplot2 will add a default scale when none is specified by a user.

Four categories of scales

• “Position scales” are used to map continuous and discrete variables onto the plotting region and construct the corresponding axes
• “Colour scales” are used to map continuous and discrete variables to colours
• “Manual scales” are used to map discrete variables to a user’s choice of symbol size, line type, shape, or colour, and create the corresponding legend
• “The identity scale” is used to plot variable values directly to the aesthetic rather than mapping them (to some other aesthetics)

Set ranges for an axis

# set ranges for both x-axis and y-axis
lims(...)

# set range for x-axis
xlim(...)

# set range for y-axis
ylim(...)

A remark

By default, the limits of position scales extend (or expand) a little past the range of data. This ensures that data do not overlap axes.

One can control the amount of expansion with the expand argument. This parameter should be a numeric vector of length two. The first element gives the multiplicative expansion, and the second the additive expansion.

If no expansion is needed, use

scale_x_continuous(expand=c(0,0))

Base plot

Plot cty (mpg in city) vs hwy (mpg on highway):

> p =ggplot(mpg)+geom_point(aes(x=cty,y=hwy,color=class)); p

Set range for x-axis

> p + xlim(c(4,20))

> # use xlim(c(NA,20)) to set an automatic lower limit

Set range for both axes

> p + lims(x = c(10, 20), y = c(3, 5))

scale_*_continuous

scale_x_continuous (or scale_y_continuous) controls the x (or y) axis for continuous variables, and often sets breaks, labels, na.value, and/or trans:

• breaks: a numeric vector of tick positions
• labels: a character vector giving labels (must be same length as breaks)
• na.value=value: missing values are set as value
• trans: transformations such as scale_*_log10(), scale_*_sqrt() and scale_*_reverse()

Illustration: base plot

> # Plot `displ` vs `hwy`:
> p1 = ggplot(mpg, aes(displ,hwy)) + geom_point(); p1

Illustration: breaks

> #  choose where the x-axis ticks appear
> p1 + scale_x_continuous(breaks = c(2, 4, 6))

Illustration: label

> #  personalized labels for ticks at specified positions
> p1 + scale_x_continuous(breaks = c(2, 4, 6),
+   label = c("two", "four", "six"))

Illustration: trans

> # y-axis on natural logarithmic scale via `trans=log`
> p1 + scale_y_continuous(trans = "log")

trans: options

Table 6.2 from book “ggplot2”

scale_*_discrete

scale_x_discrete (or scale_y_discrete):

• controls the x (or y) axis for discrete variables
• is often used to set breaks, labels, na.value, and/or trans.
• has syntax and usage similar to those of scale_x_continuous (and scale_y_continuous)

Illustration

Base layer: bar plot for drv:

> p = ggplot(mpg, aes(x = drv)) + geom_bar(); p

Illustration

> # re-label x-axis ticks
> p + scale_x_discrete(labels = 
+     c("4 wheel drive", "front drive", "rear drive"))

Overview

After position, probably the most commonly used aesthetic is colour. For this aesthetic and continuous variables, there are three methods, based on their gradient schemes:

• scale_*_gradient()
• scale_*_gradient2()
• scale_*_gradientn())

Note: colour is exchangeable with color

Method I: Two-colour gradient

scale_colour_gradient() and scale_fill_gradient():

• each being a two-colour gradient “low-high”, i.e., with a low end and a high end
• arguments low (for “low end” ) and high (for “high end”) control the colours at the low end and high end of the gradient, respectively

Method II: Diverging-colour gradient

scale_colour_gradient2() and scale_fill_gradient2():

• each being a three-colour gradient “low-mid-high”, i.e., with a low, mid, and high end
• each having a mid colour for the colour of midpoint
• midpoint defaults to \(0\) but can be set to any value

These two functions are particularly useful for creating diverging colour schemes

Method III: n-colour gradient

scale_colour_gradientn() and scale_fill_gradientn():

• each being a custom n-colour gradient
• each requiring a vector of colours in the colours argument; by default, these colours will be evenly spaced along the range of the data

Illustration

> # Plot `cty` vs `hwy` with default scheme `color=displ`
> p = ggplot(mpg)+geom_point(aes(cty,hwy,color=displ)) 
> p  # note legend title "displ""

Adjust coloring: low-mid-high

> p2a1=p + scale_colour_gradient2("Displacement",low="gray",
+       mid="blue",high="red",midpoint=mean(mpg$displ))
> p2a1 # note legend title "Displacement" and `midpoint`

Remark on midpoint

> p2a2 = p+scale_colour_gradient2("Displacement",low="gray",
+       mid="blue",high="red",midpoint=2*mean(mpg$displ))
> library(gridExtra); grid.arrange(p2a1,p2a2,nrow=2)

Visualise 3D surfaces in 2D

The faithful dataset (in library MASS) records waiting times (waiting) between eruptions and eruption times in minutes eruption for the Old Faithful geyser in Yellowstone Park:

> library(MASS); head(faithful)
  eruptions waiting
1     3.600      79
2     1.800      54
3     3.333      74
4     2.283      62
5     4.533      85
6     2.883      55

Density function \(z=f(x,y)\) for \((x,y)\)=(eruptions,waiting) can be visualized via 2D contours

Obtain estimated density

Obtain density estimate for (eruptions,waiting):

• load library MASS; apply kde2d, i.e., 2D kernel density estimation, to faithful data set
• data frame “f2d” has 3 columns x, y and z, where z is the value of the estimated density evaluated at (x,y)

> f2d <- with(faithful, MASS::kde2d(eruptions, waiting,
+ h = c(1, 10), n = 50))
> df <- with(f2d, cbind(expand.grid(x, y), as.vector(z)))
> names(df) <- c("eruptions", "waiting", "density")
> head(df)
  eruptions waiting     density
1  1.600000      43 0.003216159
2  1.671429      43 0.004146406
3  1.742857      43 0.004987802
4  1.814286      43 0.005611508
5  1.885714      43 0.005921813
6  1.957143      43 0.005882327

Visualize density estimate

> erupt <- ggplot(df,aes(waiting,eruptions,fill = density))+
+     geom_tile()+ 
+    scale_x_continuous(expand = c(0, 0))+
+    scale_y_continuous(expand = c(0, 0))

Note the use of

• geom_tile() and fill = density
• scale_*_continuous(expand = c(0, 0))

Visualize density estimate

> erupt  #max(df$density)=0.037, min(df$density)=10^(-24)

Adjust coloring: low-high

> # `limits = c(0, 0.04)` sets range for values in legend
> erupt + scale_fill_gradient(limits = c(0,0.04),
+ low = "white", high = "black")

Adjust coloring: low-mid-high

> erupt + scale_fill_gradient2(limits = c(0, 0.04),
+ midpoint = mean(df$density))

Remarks

• Pay attention to how the range for density in the legend is controlled by limits in scale_fill_gradient or scale_fill_gradient2

• There are palettes available for color scales

Overview

Two methods for colour scales for discrete data:

• choosing evenly spaced colors frm the color wheel (via, e.g., scale_colour_hue()); scale_colour_hue() works well for up to about eight colours
• selecting colors from hand-picked sets (via , e.g., RColorBrewer)

Popular palettes of RColorBrewer are “Set1” and “Dark2” for points and “Set2”, “Pastel1”, “Pastel2” and “Accent” for areas. RColorBrewer::display.brewer.all() lists all palettes.

Illustration

Part of msleep data set (from library ggplot2):

# A tibble: 6 x 3
   brainwt  bodywt vore 
     <dbl>   <dbl> <chr>
1 NA        50     carni
2  0.0155    0.48  omni 
3 NA         1.35  herbi
4  0.00029   0.019 omni 
5  0.423   600     herbi
6 NA         3.85  herbi

• brainwt (brain weight in kilograms); bodywt (body weight in kilograms); vore (carnivore, omnivore or herbivore)

Illustration

Plot brainwt vs bodywt and color “point” by vore:

> p4 = ggplot(msleep)+
+   geom_point(aes(brainwt,bodywt,colour = vore))+
+   scale_x_continuous(trans="log")+
+   scale_y_continuous(trans="log")

Note: both axes on nautral logarithmic scale

Illustration

> p4

Adjust coloring by brewer

> p4 + scale_colour_brewer(palette = "Set1")

License and session Information

License

> sessionInfo()
R version 3.5.0 (2018-04-23)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252 
[2] LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods  
[7] base     

other attached packages:
[1] MASS_7.3-49 knitr_1.21 

loaded via a namespace (and not attached):
 [1] compiler_3.5.0  magrittr_1.5    tools_3.5.0    
 [4] htmltools_0.3.6 revealjs_0.9    yaml_2.2.0     
 [7] Rcpp_1.0.0      stringi_1.2.4   rmarkdown_1.11 
[10] stringr_1.3.1   xfun_0.4        digest_0.6.18  
[13] evaluate_0.12