Skip to main content

Data Analytics Coding Fundamentals: UVic BIDA302: Course Book

22 Communication: plot formatting

22.1 Reading & reference

Garrett Grolemund and Hadley Wickham, R for Data Science

{ggplot2} reference site

22.1.1 More ggplot2 plotting resources

22.1.2 More general plotting resources

Note: this exercise does not delve into the question of how to design your plot. The structure, use of colour, annotations, and other plot elements can significantly improve the impact of a plot. See Kieran Healy’s book above, as well as

22.2 Introduction

In the chapters [data_viz] and [data_viz_2], we started to explore some ways to format our plots.

Cole Nussbaumer Knaflic’s book Storytelling with Data: A Data Visualization Guide for Business Professionals is a good introduction to the principles of data visualization, which is a key part of data analytics. In the book, the point is made that data visualization is always in the service of making a point about what the data tell us. In the context of business, this then translates into influencing decisions.

The book lists six principles (see the Storytelling with Data: quick reference guide):

  1. Understand the context

  2. Choose the right type of display

  3. Eliminate clutter

  4. Draw attention to where you want it

  5. Tell a visual story

  6. Practice makes perfect

The Storytelling with Data: quick reference guide also includes the pre-attentive attributes of a visualization, and the Gestalt principles of visual perception:

  • proximity

  • similarity

  • enclosure

  • closure

  • continuity

  • connection

One of the great things about the {ggplot2} package is that it provides virtually infinite ways to make our already-good plots even better. Here are some ways to start to incorporate formatting and design elements that improve our plots.

22.3 Text labels

Another type of annotation is to use text. For the example below, the names of the top teams are added to the 2002 season data (the year in which the Moneyball story is set). First, filter for 2002.

The next step is to identify the teams at the extremes of each quadrant—top winners and biggest losers in the above- and below-average spending teams. For this, we will use the mutate() function to add a new variable, based on the other values, using the case_when() function.

mlb_2002 <- mlb_pay_wl %>% 
  filter(year_num == 2002)

# add salary group
mlb_2002 <- mlb_2002 %>% 
  mutate(salary_grp = case_when(
    pay_index >= 100 ~ "above",
    pay_index < 100 ~ "below"
  ))


# add quadrant
mlb_2002 <- mlb_2002 %>% 
  mutate(team_quad = case_when(
    pay_index >= 100 & w_l_percent >= 0.5 ~ "I",
    pay_index < 100 & w_l_percent >= 0.5 ~ "II",
    pay_index < 100 & w_l_percent < 0.5 ~ "III",
    pay_index >= 100 & w_l_percent < 0.5 ~ "IV"
  ))

team_for_label <- mlb_2002 %>% 
  group_by(salary_grp) %>% 
  filter(w_l_percent == max(w_l_percent) |
         w_l_percent == min(w_l_percent))

team_for_label
## # A tibble: 5 × 10
## # Groups:   salary_grp [2]
##   year_num tm    attend_g est_payroll pay_index     w     l w_l_percent salary_grp team_quad
##      <dbl> <chr>    <dbl>       <dbl>     <dbl> <dbl> <dbl>       <dbl> <chr>      <chr>    
## 1     2002 CHC      33248    75690833     112.     67    95       0.414 above      IV       
## 2     2002 DET      18795    55048000      81.4    55   106       0.342 below      III      
## 3     2002 NYY      43323   125928583     186.    103    58       0.64  above      I        
## 4     2002 OAK      26788    40004167      59.2   103    59       0.636 below      II       
## 5     2002 TBD      13157    34380000      50.8    55   106       0.342 below      III

Note that we end up with five teams in the list, since Detroit (“DET”) and Tampa Bay (“TBD”) ended up with identical win-loss records.

We can now use the team names from that table as annotations, using geom_text.

# the same plot as before, but with just the 2002 teams
ggplot(mlb_2002, aes(x = pay_index, y = w_l_percent)) + 
  geom_point() +
  # add the names from the "team_for_label" table
  geom_text(data = team_for_label, aes(label = tm))

This isn’t entirely satisfactory, since the labels overlie the points. In the version below, the geom_label() is used instead, along with the nudge_x argument to move the label slightly to the left of the point (that is, -6 units on the x-axis—and yes, I experimented a bit to find the right nudge!).

# the same plot as before, but with just the 2002 teams
ggplot(mlb_2002, aes(x = pay_index, y = w_l_percent)) + 
  geom_point() +
  # add horizontal and vertical lines
  geom_vline(xintercept  = 100, colour = "grey") +
  geom_hline(yintercept = 0.5, colour = "grey") +
  # add the names from the "team_for_label" table
  geom_label(data = team_for_label, aes(label = tm),
             nudge_x = -6)

Another approach would be to omit the points altogether, and have the team abbreviations represent the location of each team on the plot:

ggplot(mlb_2002, aes(x = pay_index, y = w_l_percent)) + 
  # add horizontal and vertical lines
  geom_vline(xintercept  = 100, colour = "grey") +
  geom_hline(yintercept = 0.5, colour = "grey") +
  # plot the team names 
  geom_text(aes(label = tm))

For another example of this sort of labeling, see R for Data Science, “Graphics for Communication: Annotations”.

22.4 Annotations

The scatterplot we made in [moneyball] is interesting in and of itself. But with some annotations, some of the details can be made explicit.

One way to do that is to add lines to a plot that create sections to the plot. In the example in the previous chapter, vertical line and horizontal lines were added by using geom_vline() and geom_hline().

As you will recall, the red line runs vertically at the “1” point on the X axis. The teams to the left of the line spent below the league average for that season; the teams to the right spent more. As you can see, there have been cases when some teams spent twice as much as the league average. The blue line runs horizontally at the “0.5” point on the Y axis. Above this line, the teams won more games than they lost. Below the line, they lost more games than they won.

mlb_pay_wl <- read_csv("data/mlb_pay_wl.csv", 
                       col_types = 
                         cols(year_num = col_character()))

# create plot object for repeated use
moneyball_plot <- ggplot(mlb_pay_wl, aes(x = pay_index, y = w_l_percent)) + 
  geom_point() 

moneyball_plot +
  geom_vline(xintercept  = 100, colour = "red", size = 2) +
  geom_hline(yintercept = 0.5, colour = "blue", size = 2)

In the chart above, the vertical and horizontal lines are plotted on top of the points, obscuring some from view. Here are a couple of approaches that resolve that.

First, it’s useful to understand that the layers are plotted in order…so if we plot the lines first and then the points, the points will be in front of the lines.

ggplot(mlb_pay_wl, aes(x = pay_index, y = w_l_percent)) + 
  # plot the lines first
  geom_vline(xintercept  = 100, colour = "red", size = 2) +
  geom_hline(yintercept = 0.5, colour = "blue", size = 2) +
  # then the points
  geom_point()

A second approach is to use the alpha parameter when describing the lines. “Alpha refers to the opacity of a geom. Values of alpha range from 0 to 1, with lower values corresponding to more transparent colors.” (From the {ggplot2} reference on “Colour related aesthetics: colour, fill, and alpha”)

For our plot, we will set alpha = 0.5 for both lines.

moneyball_plot +
  geom_vline(xintercept  = 100, colour = "red", size = 2, alpha = 0.5) +
  geom_hline(yintercept = 0.5, colour = "blue", size = 2, alpha = 0.5)

Another option would have been to add a block of shading for the “Moneyball” teams that spent below the league average, but had winning seasons. For this, the annotate("rect") is used. The fill colour is hex code for the specific shade of green used by the Oakland Athletics, as found at usteamcolors.com. (See R Graphics Cookbook, 2nd ed. for applications of this in a line plot.)

moneyball_plot +
  annotate("rect", xmin = 0, xmax = 100, ymin = 0.5, ymax = 0.8,
           alpha = .3, fill = "#003831")

To that plot we might want to add a description of the quadrant that is shaded.

Note that in the label, the “” is used to force a line break in the text string.

# as previously plotted
moneyball_plot +
  annotate("rect", xmin = 0, xmax = 100, ymin = 0.5, ymax = 0.8,
           alpha = .3, fill = "#003831") +
  # text annotation
  annotate("text",             # annotation type
           label = "Teams with\nbelow-average payroll\n& winning records", 
           x = 5, y = 0.75,    # location of annotation
           hjust = 0,
           fontface = "bold",  # text formatting
           colour = "#ffffff")

22.5 Colour

Another way to tell the “Moneyball” story would be to identify the Oakland Athletics in the mass of dots shown, by using color. The code is in two steps:

  • first, create a subset of the main table that is only the Oakland data points

  • redraw the plot, adding a new geom_point() layer. Note that the colour used is now the yellow-gold in the Athletic’s colour scheme, and the size of the points is specified to be slightly larger than the others.

oakland <- mlb_pay_wl %>% 
  filter(tm == "OAK")

moneyball_plot +
  geom_point(data = oakland, aes(x = pay_index, y = w_l_percent),
             colour = "#efb21e", size = 2)

We might want to make other changes in this plot before including it in our publication.

In this version, the plot has a number of variations:

  • a light theme, using theme_bw()

  • pale gray points for all of the team points

  • green points for Oakland, using a named colour (“seagreen”)

  • adding the vertical and horizontal lines, but gray and lighter than the default

# create new version of plot with 
# gray points on white background
ggplot(mlb_pay_wl, aes(x = pay_index, y = w_l_percent)) + 
  geom_point(colour = "gray75") +
  theme_bw() +
  # Oakland points
  geom_point(data = oakland, aes(x = pay_index, y = w_l_percent),
             colour = "seagreen", size = 2) +
  geom_vline(xintercept  = 100, colour = "grey50", size = 0.75) +
  geom_hline(yintercept = 0.5, colour = "grey50", size = 0.75)

22.6 Colour palettes

There are many pre-defined colour palettes available when plotting in {ggplot2}.

(The {ggplot2} reference page that covers this is under “Scales” https://ggplot2.tidyverse.org/reference/index.html#section-scales))

An important thing to recognize (and this took me a long time to figure out!) is that how you specify the palette needs to match the type of variable that’s being represented by the colour. In general, they are either discrete (categories, such as factors or character strings) or continuous (a range of numbers).

First, let’s add “year_num” as a colour variable to a subset of our Moneyball plot.

Important: In this example, “year_num” is a specified as a “character” variable, it is therefore a discrete variable—but beware: if it had been read as a numeric variable, it would be continuous!

mlb_pay_wl %>% 
  filter(year_num > "2014") %>% 
  ggplot(aes(x = pay_index, y = w_l_percent, color = year_num)) + 
  geom_point()

We can change the default palette in a number of ways. Below, we use one of the “ColorBrewer” palettes. These palettes are designed for discrete scales on maps, but they translate well to data plotting. There are palettes for sequential, diverging, and qualitative colour scales.

A valuable reference is at R Graphics Cookbook, 2nd ed., “Using a Different Palette for a Discrete Variable”, which includes the names and images of a variety of palettes.

(See also http://colorbrewer2.org.)

And for even more information about the use of colorbrewer.org scales, see this page at the {ggplot} reference: https://ggplot2.tidyverse.org/reference/scale_brewer.html

mlb_pay_wl %>% 
  filter(year_num > "2014") %>% 
  ggplot(aes(x = pay_index, y = w_l_percent, colour = year_num)) + 
  geom_point() +
  # add ColorBrewer palette
  scale_colour_brewer(palette = "Set1")

In the plots below, the average attendance is plotted as a colour aesthetic. Average attendance is a continuous variable, so we need to make sure our palette can represent that type.

The “ColorBrewer” palettes, while designed for discrete scales, can be adapted to a continuous scale (as we have here) by using one of the distiller scales, which interpolate the values between the discrete colours in the original scale.

ggplot(mlb_pay_wl, aes(x = pay_index, y = w_l_percent, color = attend_g)) + 
  geom_point() +
  scale_color_distiller(palette = "Greens")

Another option is to use one of the viridis palettes. Note that in this case, the default scale has the largest value plotted as the lightest colour, which seems counter-intuitive to me, so I have added the direction = -1 argument to the scale_color_viridis_c() function.

ggplot(mlb_pay_wl, aes(x = pay_index, y = w_l_percent, color = attend_g)) + 
  geom_point() +
  scale_color_viridis_c(direction = -1)

For more information about the viridis palettes, see R Graphics Cookbook, 2nd ed, “Using Colors in Plots: Using a Colorblind-Friendly Palette”

22.7 Animated plots

If you are interested in exploring animated plots (similar to Hans Rosling’s gapminder visualizations), here are some resources for using the {gganimate} package:

-30-