The ggplot package

It is the most elegant and versatile tool for graphically visualizing data in R, offering a coherent system (or grammar) for building graphs.

What is a grammar of graphics?

• It is a concept coined by Leland Wilkinson in 2005.

• An abstraction which facilitates reasoning and communicating graphics.

• ggplot2 is a layered grammar of graphics which allow users to: independently specify the building blocks of a plot and combine them to create just about any kind of graphical display.

ggplot2 characteristics

Advantages of ggplot2:

• The package is flexible and offers extensive customization options.
• The documentation is well-written.
• ggplot2 has a large user base => it’s easy find to help.

Weaknesses of ggplot2

• it does not handle 3D graphics
  • use rgl or ggplot2 + plotly instead,
• it is inefficient for graph/network plots with nodes and edges
  • use igraph instead
• does not offer interactive graphics:
  • use ggvis, or plotly instead

Building blocks of a ggplot2 graphical objects

ggplot(data = <DATA>) +
  GEOM_FUNCTION(
    mapping = aes(<mappings>),
    stat = <statistic transformation>,
    position = <position options>,
    color = <fixed color>, 
    <other arguments>) +
  FACET_FUNCTION(<facet options>) +
  SCALE_FUNCTION(<scale options>) +
  theme(<theme elements>)

ggplot() function

• ggplot() function is initializes a basic graph structure.

• It cannot produce a plot alone by itself.

• You need to add extra components to generate a graph.

• Different parts of a plot can be added together using +. Note similarity with the %>% operator.

• Any data or arguments you supply to ggplot() function, can later be used by added functions without repeated specification.

ggplot2 compared to base graphics

• is more verbose for simple/out of the box graphics,

• is less verbose for complex/custom graphics,

• generates graphs by adding building blocks, instead of calling different functions to draw new layers on top,

• makes it easier to edit and tweak elements of a plot,

• more details on advantages of ggplot2 over base plot are in this blog.

Example 1: History of unemployment

ggplot2 has a built-in economics dataset, which inclides time series data on US unemployment from 1967 to 2015.

economics

## # A tibble: 574 x 6
##    date         pce    pop psavert uempmed unemploy
##    <date>     <dbl>  <int>   <dbl>   <dbl>    <int>
##  1 1967-07-01  507. 198712    12.5     4.5     2944
##  2 1967-08-01  510. 198911    12.5     4.7     2945
##  3 1967-09-01  516. 199113    11.7     4.6     2958
##  4 1967-10-01  513. 199311    12.5     4.9     3143
##  5 1967-11-01  518. 199498    12.5     4.7     3066
##  6 1967-12-01  526. 199657    12.1     4.8     3018
##  7 1968-01-01  532. 199808    11.7     5.1     2878
##  8 1968-02-01  534. 199920    12.2     4.5     3001
##  9 1968-03-01  545. 200056    11.6     4.1     2877
## 10 1968-04-01  545. 200208    12.2     4.6     2709
## # ... with 564 more rows

economics <- mutate(economics, unemp_rate = unemploy/pop)

R base graphics

plot(unemp_rate ~ date, data = economics,  type = "l")

ggplot2 package

library(tidyverse)
ggplot(data = economics, aes(x = date, y = unemp_rate)) + geom_line()

ggplot() by itself does not plot the data

ggplot(data = economics, aes(x = date, y = unemp_rate))

You need to add a line-layer

ggplot(data = economics, aes(x = date, y = unemp_rate)) + geom_line()

Change the background color to white

ggplot(data = economics, aes(x = date, y = unemp_rate)) + 
  geom_line() + theme_bw()

What about comparing 2009 to 2014?

# Add new variables for plotting
economics <- economics %>% 
  mutate(month = as.numeric(format(date, format="%m")),
         year  = as.factor(format(date, format="%Y")))
economics %>% 
  select(date, month, year, unemp_rate)

## # A tibble: 574 x 4
##    date       month year  unemp_rate
##    <date>     <dbl> <fct>      <dbl>
##  1 1967-07-01     7 1967      0.0148
##  2 1967-08-01     8 1967      0.0148
##  3 1967-09-01     9 1967      0.0149
##  4 1967-10-01    10 1967      0.0158
##  5 1967-11-01    11 1967      0.0154
##  6 1967-12-01    12 1967      0.0151
##  7 1968-01-01     1 1968      0.0144
##  8 1968-02-01     2 1968      0.0150
##  9 1968-03-01     3 1968      0.0144
## 10 1968-04-01     4 1968      0.0135
## # ... with 564 more rows

Using base graphics

data09 <- subset(economics, year == "2009")
data14 <- subset(economics, year == "2014")
plot(unemp_rate ~ month, data = data09, ylim = c(0.02, 0.05), type = "l")
lines(unemp_rate ~ month, data = data14, col = "red")
legend("topleft", c("2009", "2014"), col = c("black", "red"), lty = c(1,1))

Using ggplot2

There is no need of specifing a legend:

ggplot(data = economics %>% filter(year %in% c(2014, 2009)), 
       aes(x = month, y = unemp_rate)) + 
  geom_line(aes(group = year, color = year)) 

Plotting all the years together is easy

ggplot(data = economics, aes(x = month, y = unemp_rate)) + 
  geom_line(aes(color = year)) +
  theme(axis.text.x = element_text(angle = 45))

Plotting all the years together is easy

ggplot(data = economics, aes(x = month, y = unemp_rate)) + 
  geom_line(aes(group = year, color = pop)) +
  theme(axis.text.x = element_text(angle = 45))

The diamond dataset

diamond is a built-in dataset, included in tidyverse. It contains prices and other attributes of almost 54,000 diamonds. We will use this dataset to illustrate how to use functions in ggplot2.

data(diamonds)
diamonds

## # A tibble: 53,940 x 10
##    carat cut       color clarity depth table price     x     y     z
##    <dbl> <ord>     <ord> <ord>   <dbl> <dbl> <int> <dbl> <dbl> <dbl>
##  1 0.23  Ideal     E     SI2      61.5    55   326  3.95  3.98  2.43
##  2 0.21  Premium   E     SI1      59.8    61   326  3.89  3.84  2.31
##  3 0.23  Good      E     VS1      56.9    65   327  4.05  4.07  2.31
##  4 0.290 Premium   I     VS2      62.4    58   334  4.2   4.23  2.63
##  5 0.31  Good      J     SI2      63.3    58   335  4.34  4.35  2.75
##  6 0.24  Very Good J     VVS2     62.8    57   336  3.94  3.96  2.48
##  7 0.24  Very Good I     VVS1     62.3    57   336  3.95  3.98  2.47
##  8 0.26  Very Good H     SI1      61.9    55   337  4.07  4.11  2.53
##  9 0.22  Fair      E     VS2      65.1    61   337  3.87  3.78  2.49
## 10 0.23  Very Good H     VS1      59.4    61   338  4     4.05  2.39
## # ... with 53,930 more rows

More information with ?diamonds. Spreadsheet view in RStudio with View(diamonds).

Geometic object

Geometric objects are the actual elements you put on the plot. Examples include:

• points (geom_point(), used for scatter plots)
• text (geom_text(), geom_label(), used for text labels)
• lines (geom_line(), used for time series, trend lines, etc.)
• boxplots (geom_boxplot() used for, well, boxplots!)

There is no upper limit to how many geom objects you can use. You can add a geom objects to a plot using an + operator.

To get a list of available geometric objects use the following:

help.search("geom_", package = "ggplot2")

Scatter plots

# Note that we can save `ggplot` as an object
p <- ggplot(diamonds, aes(x = carat, y = price))
p + geom_point()

Text labels plots

plog <- ggplot(
  sample_n(diamonds, 100), 
  aes(x = log10(carat), y = log10(price)))
plog + geom_text(aes(label = clarity))

Text plots with rectangle plates

plog + geom_label(aes(label = clarity))

ggrepel package for annotation

ggrepel helps annotating overlapping labels.

# Uncomment the line below if you don't have 'ggrepel'
# install.packages("ggrepel")
library(ggrepel)
plog + geom_point() + geom_text_repel(aes(label = clarity), size = 3)

Aesthetic mapping

• In ggplot an aesthetic mapping, defined with aes(), describes how variables are mapped to visual properties or aesthetics.

• The details of mapping can be described by using scale functions.

• Aesthetics are properties you can see:
  • position (i.e., on the x and y axes)
  • shape
  • linetype
  • size
  • color (“outside” color)
  • fill (“inside” color)

You can convey information about your data by mapping the aesthetics in your plot to the variables in your dataset.

Each type of geom objects accepts only a subset of aesthetics; refer to the geom help pages for details.

The shape of the points

# We first generate a subset of 'diamnonds' dataset
dsmall <- sample_n(diamonds, 500)
p1 <- ggplot(dsmall, aes(x = carat, y = price))

# set shape by diamond cut
p1 + geom_point(aes(shape = cut))

## Warning: Using shapes for an ordinal variable is not advised

All 25 shape configurations

ggplot(data.frame(x = 1:5 , y = 1:25, z = 1:25), aes(x = x, y = y)) +
  geom_point(aes(shape = z), size = 5, colour = "darkgreen", fill = "orange") +
  scale_shape_identity()

The color of the points

# color by diamonds color
p1 + geom_point(aes(color = color))

Set color and shape

p1 + geom_point(aes(shape = cut, color = color))

## Warning: Using shapes for an ordinal variable is not advised

Variable vs fixed aesthetics

p1 + geom_point(aes(color = color))

p1 + geom_point(color = "darkgreen") 

Marker points with borders

p1 + geom_point(aes(fill = cut), size = 3, color = "black", shape = 25)

Alpha parameter for transparency

a1 <- p + geom_point(alpha = 1/5)
a2 <- p + geom_point(alpha = 1/50)
a3 <- p + geom_point(alpha = 1/500)

# We use grid.arrange from gridExtra to display multiple plots
library(gridExtra)
grid.arrange(a1, a2, a3, ncol = 3)

Aesthetic mapping vs variable scaling

• aes() assign an aesthetic to a variable; it doesn’t determine how mapping should be done.

• For example, aes(shape = x) or aes(color = z) do not specify what shapes or what colors should be used.

• To choose colors/shapes/sizes etc. you need to modify the corresponding scale.

• ggplot2 includes scales for:
  • position
  • color and fill
  • size
  • shape
  • line type

• Scales can be modified with functions of the form: scale_<aesthetic>_<type>(), e.g. scale_color_discrete().
  

• Try typing scale_<tab>() to see a list of scale modification functions.
  

• Common Scale Arguments:

  • name: the first argument gives the axis or legend title
  • limits: the minimum and maximum of the scale
  • breaks: the points along the scale where labels should appear
  • labels: the labels that appear at each break

Scales for the axes

# Square root y-axis transformation
p1 <- ggplot(dsmall, aes(x = carat, y = price)) 
psqrt <- p1 + geom_point() + scale_y_sqrt()
# Log base 10 y-axis transformation
plog10 <- p1 + geom_point() + scale_y_log10()
grid.arrange(psqrt, plog10, ncol = 2)

Log base 10 transformation of x and y axes. Note the differences.

ploglog1 <- p1 + geom_point() + scale_y_log10() + scale_x_log10()
ploglog2 <- ggplot(dsmall, aes(x = log(carat), y = log(price))) + geom_point()
grid.arrange(ploglog1, ploglog2, ncol = 2)

Scales for shapes

p11 <- p1 + geom_point(aes(shape = cut), size = 3)  
p12 <- p1 + geom_point(aes(shape = cut), size = 3) + 
  scale_shape_manual(values = c(1:5))
grid.arrange(p11, p12, ncol = 2)

## Warning: Using shapes for an ordinal variable is not advised

Scales for colors

To choose specific colors for discrete variables we use scale_color_manual.

p1 + geom_point(aes(color = cut), size = 3) + 
  scale_color_manual(values = c("red", "orange", "yellow", "green", "blue"))

For continuous variables we use scale_color_gradient, and specify the ends of the color spectrum.

p1 + geom_point(aes(color = price), size = 3) + 
  scale_color_gradient(low = "blue", high = "red")

You can also scale the values of the variable corresponding to color.

p1 + geom_point(aes(color = price), size = 3) + 
  scale_color_gradient(low = "blue", high = "red", trans = "log10")

Or and add your own breaks

p1 + geom_point(aes(color = price), size = 3) + 
  scale_color_gradient(low = "blue", high = "red", trans = "log10",
                       breaks = c(1000, 2000, 5000, 10000),
                       labels = c("  1000", "  2000", "  5000", "10000")) 

scale_color_brewer lets you choose nice color palettes for discrete variables.

p1 + geom_point(aes(color = cut), size = 3) + 
  scale_color_brewer(palette = "Set2")

Unfortunately, scale_color_brewer doesn’t work for continuous variables:

# This will result in an error
p1 + geom_point(aes(shape = price), size = 3) + 
  scale_color_brewer(palette = "Spectral")

## Error: A continuous variable can not be mapped to shape

We can get around this issue using the RColorBrewer package and scale_color_gradientn function, which interpolates colors from the brewer palettes.

# install.packages("RColorBrewer")
library(RColorBrewer)
p1 + geom_point(aes(color = price), size = 3) + 
  scale_color_gradientn(colours = brewer.pal(name = "Spectral", n = 10))

Another popular color scheme package, viridis, supports both discrete and continuous variables:

# install.packages("viridis")
library(viridis)
p1 + geom_point(aes(color = price), size = 3) + scale_color_viridis()

… there are also other unconventional schemes such as, one based on Wes Anderson movies :

#install.packages("wesanderson")
library(wesanderson)
names(wes_palettes)

##  [1] "BottleRocket1"  "BottleRocket2"  "Rushmore1"      "Rushmore"      
##  [5] "Royal1"         "Royal2"         "Zissou1"        "Darjeeling1"   
##  [9] "Darjeeling2"    "Chevalier1"     "FantasticFox1"  "Moonrise1"     
## [13] "Moonrise2"      "Moonrise3"      "Cavalcanti1"    "GrandBudapest1"
## [17] "GrandBudapest2" "IsleofDogs1"    "IsleofDogs2"

Wes Anderson color palette:

# For discrete variables
p1 + geom_point(aes(color = cut), size = 3) + 
  scale_color_manual(values = wes_palette("Darjeeling1", n = 5))

Facettng allows you to split up your data by one or more variables and plot the subsets of data together.

dsmall <- diamonds[sample(nrow(diamonds), 1000), ]
p0 <- ggplot(data = dsmall, aes(x = carat, y = price)) +geom_point(size = 1) +
  geom_smooth(aes(colour = cut, fill = cut)) 
p1 <- p0 + facet_wrap(~ cut)
grid.arrange(p0, p1, ncol = 2)

Exercise 1

• Go to “Lec4_Exercises.Rmd” on the class website.

• Complete Exercise 1.

Types of statistical transformations

Plots often require some statistical data transformation or computation
before they can be plotted. Examples include:

• boxplots: the the median, lower and upper quartiles,

• histograms: group the values into bins,

• bar charts: number of class occurrences.

• smoothers: prediction lines / predicted y-values,

Box plot transformation

Plotting a summary (less data) can be more insightful.

ggplot(data = diamonds, aes(x = cut, y =carat)) +
  geom_boxplot()

Histogram and density plots

# Distribution of the carats (weights) of the diamonds.
h <- ggplot(data = diamonds, aes(x = carat)) + geom_histogram()
d <- ggplot(data = diamonds, aes(x = carat)) + geom_density()
grid.arrange(h, d, ncol = 2)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Histogram parameters

In histograms, the smoothness is controlled with bins and binwidth arguments. (or by specifying using the breaks explicitly).

p <- ggplot(data = diamonds, aes(x = carat)) + xlim(0, 3)
h1 <- p + geom_histogram(binwidth = 0.5) 
h2 <- p + geom_histogram(binwidth = 0.1) 
h3 <- p + geom_histogram(binwidth = 0.05) 
grid.arrange(h1, h2, h3, ncol = 3)

Density plot parameters

In density plots, the bw (the smoothing bandwidth) and adjust arguments control the smoothness.

d1 <- p + geom_density(adjust = 5) 
d2 <- p + geom_density(adjust = 1) 
d3 <- p + geom_density(adjust = 1/5) 
grid.arrange(d1, d2, d3, ncol = 3)

Histograms for separate groups

# Here we show grouping by diamonds cut.
h <- p + geom_histogram(aes(fill = cut), position = "dodge", bins = 10)
d <- p + geom_density(aes(color = cut))
grid.arrange(h, d, ncol = 2)

Instead of marginal distributions, we can plot distribution of components stacked on top of each other to see the contribution from each of group.

h <- p + geom_histogram(aes(fill = cut), position = "stack")
d <- p + geom_density(aes(fill = cut), position = "stack")
grid.arrange(h, d, ncol = 2)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Position adjustments

Position adjustments are used to adjust the position of each geom. The following position adjustments are available:

• position_identity: default of most geoms
• position_jitter: adds a small amount of random variation
• position_dodge: default of geom_boxplot
• position_stack: default of geom_bar, geom_histogram
• position_fill: useful for geom_bar, geom_histogram

The position parameter can be set as follows:

geom_point(..., position="jitter")

Position adjustments for scatterplots

Overplotting: many points overlap each other. Here variables are categorical, but sometimes rounding causes overplotting.

plt <- ggplot(diamonds, aes(x = cut, y = depth))
plt + geom_point()

plt + geom_point(position = "jitter")

Bar charts

• A discrete analogue of a histogram is the bar chart, geom_bar().

• Instead of partitioning the values into bins like histograms, the bar geom counts the number of instances of each discrete class. The counts are then plotted as columns for each distinct class.

• If you’d like include unequal weights for different observations, you can use the weight aesthetic.

b1 <- ggplot(diamonds, aes(x = clarity)) + geom_bar()
b2 <- ggplot(diamonds, aes(x = clarity)) + geom_bar(aes(weight = carat)) + ylab("carat")
grid.arrange(b1, b2, ncol = 2)

The left plot shows the number of diamonds in each clarity group, and the right plot shows the count weighted by carat, which is equivalent to showing the total weight of diamonds in clarity color group.

• As you see, in ggplot2 (unlike base graphics) it is not necessary tabulate the values, i.e. compute the counts of each category beforehand. The computation is done automatically for you.

• However, if you have already summarized data, you can still use geom_bar but you need to specify an identity transformation, stat = "identity rather than the default stat = "count".

diamond.counts <- diamonds %>%
    group_by(color) %>% 
    summarise(count = n())
diamond.counts

## # A tibble: 7 x 2
##   color count
##   <ord> <int>
## 1 D      6775
## 2 E      9797
## 3 F      9542
## 4 G     11292
## 5 H      8304
## 6 I      5422
## 7 J      2808

With the frequency counts already computed, the default options of the barplot generates an error:

diamond.counts

## # A tibble: 7 x 2
##   color count
##   <ord> <int>
## 1 D      6775
## 2 E      9797
## 3 F      9542
## 4 G     11292
## 5 H      8304
## 6 I      5422
## 7 J      2808

ggplot(diamond.counts, aes(x=color, y=count)) +  geom_bar()

## Error: stat_count() must not be used with a y aesthetic.

Smoothers and trend lines

# Smoothers help discern patterns in the data
set.seed(438756)
dsmall <- diamonds %>% sample_frac(0.1)
ggplot(dsmall, aes(x = carat, y = price)) + 
    geom_point(aes(color = color)) + geom_smooth()

## `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'

Regression lines with ggplot2

ggplot(dsmall, aes(x = carat, y = price)) + 
    geom_point(aes(color = color)) + geom_smooth(method = "lm")

Saving plots

Now that you have your beautiful plot, you may want to save it as an image.

ggsave() is a convenient function for saving a plot.

By default, it saves the last plot that you displayed, using the size of the current graphics device. It also guesses the type of graphics device from the extension.

ggsave(filename, plot = last_plot(), device = NULL, path = NULL,
  scale = 1, width = NA, height = NA, units = c("in", "cm", "mm"),
  dpi = 300, limitsize = TRUE, ...)

“Device” can be either be a device function (e.g. png), or one of “eps”, “ps”, “tex” (pictex), “pdf”, “jpeg”, “tiff”, “png”, “bmp”, “svg” or “wmf” (windows only).

Exercise 2, 3

• Go back to “Lec4_Exercises.Rmd”

• Complete the exercise 2 and 3