3 Visualizations

3.4.1 Basics

When using ggplot2, there are unlimited possibilities on what you can manipulate/influence. This may be daunting, but always remember that all plots work on the same framework.

plot = data + aesthetics + geometry

No matter what ggplot you are making, no matter how many characteristics you influence, all ggplot2 needs are three things:

1. The data: the data being used to make the plot
2. The aesthetics: x/y/color/shape/etc. In ggplot2 aesthetics is shortened to aes.
3. The geometry: plot type (e.g., scatterplot, boxplot, etc.)

With only those three things, you can make any type of visualization you want or need. From there, you can build as far as your mind can see. When you do want to add more levels to your plot, you do so by using the + sign.

Here is an example of a basic graph made with ggplot:

ggplot(data = mpg, aes(x = cty, y = hwy)) +
  geom_point()

Figure 3.3: An example of a visualization made using ggplot2.

This is our first ggplot (a scatterplot) we have created, so let’s break this down:

• The data: we are using the mpg dataset.
• The aesthetics: the x variable is cty and the y is hwy.
• The geometry: geom_point(), which creates points on the graph.
  • Importantly, to add geometry, you do need to add the + sign.

With only two lines of code, a scatterplot was created! Since the framework has been established, it is time to build some visualizations!

Note: In this chapter, there will be basic examples of each visualization type, and advanced examples of each visualization type. This is done to display the scope of what can be done with this package (ggplot2). It is encouraged to experiment with the code, change numbers, remove things, and compare and contrast the differences between the experimented visuals.

3.4.2 Scatterplot - geom_point()

When data has two continuous (for example, numeric) variables, and you want to visualize their relationship, a scatterplot is a fantastic choice. This is the basis for relationship analysis that topics such as linear regression (more about this in Chapter 8) rely on.

The geometry that needs to be specified is geom_point().

ggplot(data = mpg, aes(x = cty, y = hwy)) +
  geom_point()

Figure 3.4: A basic example of a scatterplot using ggplot2.

Typically it is best practice to have a “line of best fit” when creating scatterplots, similar to Anscombe’s visualization. To do that, you can utilize the geom_smooth() command.

# Line of Best Fit
ggplot(mpg, aes(x = cty, y = hwy)) +
  geom_point() +
  geom_smooth(method = "lm") # lm stands for "linear model"
#> `geom_smooth()` using formula = 'y ~ x'

Figure 3.5: An example of a scatterplot with a line of best fit.

Notice that to add geom_smooth, which is another layer of our visualization, we needed to add another + sign. It is used quite literally in ggplot2.

Let’s add some more information to our scatterplot.

ggplot(data = mpg, aes(x = cty, y = hwy, color = class, shape = drv)) +
  geom_point(alpha = 0.8, size = 2.5) + # opacity and size
  labs(
    title = "City vs Highway MPG",      # adds a title to the plot
    x = "City MPG",                     # adds a x-axis label to the plot
    y = "Highway MPG",                  # adds a y-axis label to the plot
    color = "Vehicle Class",            # adds a color label to the plot
    shape = "Drivetrain"                # adds a shape label to the plot
  ) + 
  facet_wrap(~ class) +                 # breaks the graph into individual graphs
  geom_smooth(method = "lm", se = TRUE) # adds a line of best fit
#> `geom_smooth()` using formula = 'y ~ x'

Figure 3.6: A scatterplot incorporating multiple aesthetics and faceting.

3.4.3 Bar Chart (counts) and Column Chart (values)

Bar charts visualize counts of a discrete variable, while column charts visualize pre-summarized numeric values.

3.4.3.1 Bar Chart - geom_bar()

For a bar chart, the geometry used is geom_bar().

# BASIC (counts by class) - geom_bar() counts rows automatically
ggplot(mpg, aes(x = class)) +
  geom_bar()

Figure 3.7: An example of a basic bar chart.

Depending on how long the variable names are, it may be best to switch the x and y axis. They would still act the same, but they would just flip on the coordinate plane. The x variable would still be the x variable, and the y would still be the y variable, but flipped. To do this, you can utilize the coord_flip() command.

ggplot(mpg, aes(x = manufacturer)) +
  geom_bar(fill = "steelblue", color = "white") +
  coord_flip() +
  labs(title = "Counts by Manufacturer", x = "", y = "Count")

Figure 3.8: An example of a coordinate flipped bar chart.

Depending on the audience, a stacked bar chart may be the best way to visualize the data. To do that, you can add position = "stack" and ggplot does the rest.

# What if we want a stacked bar chart (default with fill)?
ggplot(mpg, aes(x = manufacturer, fill = drv)) +
  geom_bar(position = "stack", color = "white") +
  coord_flip() +
  labs(
    title = "Counts by Manufacturer and Drivetrain",
    x = "",
    y = "Count",
    fill = "Drivetrain"
  ) +
  theme_minimal()

Figure 3.9: An example of a stacked bar chart.

If you need to create a grouped bar chart instead, you can add position = "dodge" to create your visual.

#What if we want a grouped bar chart
ggplot(mpg, aes(x = manufacturer, fill = drv)) +
  geom_bar(position = "dodge", color = "white") +
  coord_flip() +
  labs(
    title = "Counts by Manufacturer and Drivetrain",
    x = "",
    y = "Count",
    fill = "Drivetrain"
  ) +
  theme_minimal()

Figure 3.10: An example of a grouped bar chart.

3.4.3.2 Column Chart - geom_col()

Column charts work best when you have pre-summarized values, and not raw values.

# USING PRE-SUMMARIZED VALUES - geom_col() requires explicit values
class_counts <- mpg %>%
  count(class)  # counts rows by class

kable(class_counts, caption = "Pre-summarized values")

Table 3.6: Pre-summarized values

class	n
2seater	5
compact	47
midsize	41
minivan	11
pickup	33
subcompact	35
suv	62

Once the values have been summarized explicitly, you use geom_col() to create a column chart.

ggplot(class_counts, aes(x = class, y = n)) +
  geom_col()

Figure 3.11: An example of a basic column chart.

There are a few things you can do to a column chart to add some flavor. For example:

• Use the reorder command to reorder the columns into ascending or descending order based on their n values.
• Inside of geom_col change the width of the columns.
• Depending on the bars, you can change the legend.position to a particular spot (or remove it entirely) from the visualization.

# PLUS AESTHETICS (polished)
ggplot(class_counts, aes(x = reorder(class, n), y = n, fill = class)) +
  geom_col(width = 0.7, color = "white") +   # width = bar thickness, color = border
  coord_flip() +                             # flip for readability
  labs(title = "Counts by Vehicle Class", x = "", y = "Count") +
  theme(legend.position = "none")

Figure 3.12: An example of a column chart with polished aesthetics.

3.4.4 Histograms and Density Plots (distribution)

Histograms are perfect for when you are looking to display distribution.

3.4.4.1 Histograms - geom_histogram()

For the geometry of a histogram, you use geom_histogram().

# BASIC
ggplot(mpg, aes(x = hwy)) +
  geom_histogram(binwidth = 3)

Figure 3.13: An example of a basic histogram.

R automatically assigns bins when creating a histogram unless otherwise instructed. There are several things that can be influenced:

• bin width: use binwidth to change the width of the bins
• boundary: use boundary to set the separation between the bins

# STYLED (bin edges + colors)
ggplot(mpg, aes(hwy)) +
  geom_histogram(binwidth = 5, boundary = 0,
                 fill = "red", color = "orange") +
  labs(title = "Highway MPG Distribution", x = "Highway MPG", y = "Frequency")

Figure 3.14: An example of a styled histogram.

In the case that you need a stacked histogram, below is code to create that. The secret here is using the fill command.

# MAPPED FILL (stacked by class)
ggplot(mpg, aes(hwy, fill = class)) +
  geom_histogram(binwidth = 10, color = "white") +
  labs(title = "Highway MPG Distribution by Vehicle Class",
       x = "Highway MPG", y = "Count", fill = "Vehicle Class") +
  theme(legend.position = "bottom")

Figure 3.15: An example of a filled histogram.

3.4.4.2 Density - geom_density()

Density plots still show the distribution of data, but instead of doing it in bins like a histogram, accomplish this through an outline. The geometry for a density plot is geom_density().

# BASIC
ggplot(diamonds, aes(x = price)) +
  geom_density()

Figure 3.16: An example of a basic density plot.

In the below example, the visualization is filtering within the data portion to only keep the cuts “Good”, “Ideal” and “Premium.” Since it is utilizing color inside the aesthetics, this will create a grouped density plot, showing three different lines for each of the three different cuts.

Note: If there was no filtering, this code would still separate into different lines.

# GROUPED
ggplot(diamonds %>% filter(cut %in% c("Good", "Ideal", "Premium")),
       aes(price, color = cut)) +
  geom_density() +
  labs(title = "Price Density by Cut", x = "Price", y = "Density") +
  theme(legend.position = "bottom")

Figure 3.17: An example of a grouped density plot.

3.4.5 Boxplot - geom_boxplot()

Commonly known as a box and whisker plot, boxplots are fantastic for providing numerical insights between categorical variables. There are a few different pieces of a boxplot:

• Whiskers: there are two whiskers on each boxplot
  • Lower Whisker: shows the lower 25% of the data. The bottom is the lowest value in the dataset
  • Upper Whisker: shows the upper 25% of the data. The top is the highest value in the dataset
• Box: the box itself shows the middle 50% of the data. This includes:
  • Interquartile Range (IQR): the lowest line in the bar is the 25th percentile and the top is the 75th percentile
  • Median: the darker line inside of the box
• Outliers: any data point that is above or below the upper and lower whisker, respectively.

The geometry for a boxplot is geom_boxplot().

# BASIC
ggplot(mpg, aes(x = class, y = hwy)) +
  geom_boxplot() +
  labs(title = "Highway MPG by Vehicle Class", x = "Vehicle Class", y = "Highway MPG")

Figure 3.18: An example of a basic boxplot.

Sometimes the points on a boxplot (or any plot) can be indistinguishable due to them being so close together. In that case, utilize the geom_jitter() command.

# WITH JITTERED POINTS OVERLAID
ggplot(mpg, aes(class, hwy)) +
  geom_boxplot(outlier.shape = NA) +
  geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
  coord_flip() +
  labs(title = "Highway MPG by Vehicle Class (with Points)",
       x = "", y = "Highway MPG")

Figure 3.19: An example of a boxplot with jittered points.

3.4.6 Lines (time series) - geom_line()

Time and time again, when working with time-series data, a line graph is created. The geometry for a line graph is geom_line().

# BASIC: unemployment over time
ggplot(economics, aes(x = date, y = unemploy)) +
  geom_line() +
  labs(title = "US Unemployment Over Time",
       x = "Date", y = "Number Unemployed (thousands)")

Figure 3.20: An example of a basic line graph.

Like in the scatterplot, you can add a line of best fit.

# PLUS: LM vs LOESS contrast
ggplot(economics, aes(date, unemploy)) +
  geom_line(linewidth = 1.6) +
  geom_smooth(method = "lm", se = FALSE, color = "red") +
  labs(title = "US Unemployment with Linear Trend",
       x = "Date", y = "Number Unemployed (thousands)")
#> `geom_smooth()` using formula = 'y ~ x'

Figure 3.21: An example of line graph with a LM line of best fit.

Instead of using “lm” for our method, let’s try “loess” and see our results.

ggplot(economics, aes(date, unemploy)) +
  geom_line(linewidth = 0.6) +
  geom_smooth(method = "loess", se = TRUE) +  # loess = flexible smoothing; se = confidence band
  labs(title = "US Unemployment with LOESS Smooth",
       x = "Date", y = "Number Unemployed (thousands)")
#> `geom_smooth()` using formula = 'y ~ x'

Figure 3.22: An example of line graph with a LM line of best fit.

3.4.7 Put text on the plot - geom_text()

No matter the type of plot, it may help the viewer understand your plot better if you add labels to some of your data points, for example, the most extreme. To do this, utilize the geom_text() command.

# BASIC: label extreme points
mpg_extremes <- mpg %>% slice_max(order_by = hwy, n = 5)
mpg_extremes
#> # A tibble: 6 × 11
#>   manufacturer model     displ  year   cyl trans drv     cty
#>   <chr>        <chr>     <dbl> <int> <int> <chr> <chr> <int>
#> 1 volkswagen   jetta       1.9  1999     4 manu… f        33
#> 2 volkswagen   new beet…   1.9  1999     4 manu… f        35
#> 3 volkswagen   new beet…   1.9  1999     4 auto… f        29
#> 4 toyota       corolla     1.8  2008     4 manu… f        28
#> 5 honda        civic       1.8  2008     4 auto… f        25
#> 6 honda        civic       1.8  2008     4 auto… f        24
#> # ℹ 3 more variables: hwy <int>, fl <chr>, class <chr>
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  geom_text(data = mpg_extremes, aes(label = model), nudge_y = 1, size = 3) +
  labs(title = "Top Highway MPG Models Labeled",
       x = "Engine Displacement (L)", y = "Highway MPG")

Figure 3.23: An example of adding text inside a plot.

The code above first identifies the top three hwy values using slice_max(), pairs that with geom_text() and labels only the extreme data points. Now, viewers can read from the plot which model of cars have the best highway miles per gallon.

3.4.8 Error bars (requires summary stats) - geom_errorbar()

In times where you want error bars to be displayed, you first need to compute the mean and standard error for your values.

# Compute mean & standard error for hwy by class
summ_hwy <- mpg %>%
  group_by(class) %>%
  summarize(
    mean_hwy = mean(hwy, na.rm = TRUE),
    se_hwy   = sd(hwy, na.rm = TRUE) / sqrt(n()))

Once completed, you can utilize the geom_errorbar() command to add error bars to your plot.

# Points + error bars (plot error bars first so points sit on top)
ggplot(summ_hwy, aes(class, mean_hwy)) +
  geom_errorbar(aes(ymin = mean_hwy - se_hwy, ymax = mean_hwy + se_hwy), width = 0.2) +
  geom_point(size = 2) +
  coord_flip() +
  labs(title = "Mean Highway MPG (± SE) by Class", x = "", y = "Mean Highway MPG")

Figure 3.24: An example of a plot with error bars.

3.4.9 Reference lines

Let’s say there is a scenario where you are looking for data above and below a particular threshold. In this case, reference lines can become an essential tool for yourself and your viewers. Once the threshold is established (mean, median, really any number of significance to you), you can utilize the geom_hline() or the geom_vline() commands to create horizontal or vertical reference lines, respectively.

# Horizontal line at overall mean
overall_mean <- mean(mpg$hwy, na.rm = TRUE)

ggplot(mpg, aes(displ, hwy)) +
  geom_point(alpha = 0.6) +
  geom_hline(yintercept = overall_mean, linetype = "dashed") +
  labs(title = "Reference Line at Overall Mean Highway MPG",
       x = "Engine Displacement (L)", y = "Highway MPG")

Figure 3.25: An example of plot with a horizontal reference line.

# Vertical line at displ = 3
ggplot(mpg, aes(displ, hwy)) +
  geom_point(alpha = 0.6) +
  geom_vline(xintercept = 3, linetype = "dotted") +
  labs(title = "Reference Line at Engine Displacement = 3L",
       x = "Engine Displacement (L)", y = "Highway MPG")

Figure 3.26: An example of plot with a vertical reference line.

3.7.1 Summary of ggplot Geometries

Below is a list of plot types, their purpose, and the geom command used:

• Scatterplot - Relationships - geom_point()
• Bar Chart - Counts - geom_bar()
• Column Chart - Pre-summarized values - geom_col()
• Histogram - Distribution - geom_histogram()
• Density Plot - Distribution - geom_density()
• Boxplot - Group comparison - geom_boxplot()
• Line Graph - Time - geom_line()

3.7.2 Summary of other ggplot commands

Below is a list of other commands used to alter plots:

• Aesthetics:
  • color: Changes the color of the points.
  • shape: Changes the shape of the points.
  • alpha: Changes the opacity of the point.
  • size: Changes the size of the point.
  • fill: Controls the interior color of shapes
• labs(): Creates labels, including title, x-axis, and y-axis.
• facet_wrap(): Creates individual plots and puts it into one graphic.
• coord_flip(): Flips the axes without changing the underlying variables.
• theme(): Controls the overall appearance of the plot
• theme_minimal(): Makes the most basic looking plot.
• theme(legend.position = "..."): Dictates where (if at all) the legend appears on the plot.
• geom_text(): Adds text to the data points within the plot.
• geom_hline(): Adds a horizontal reference line.
• geom_vline(): Adds a vertical reference line.
• geom_smooth(): Adds trend lines.
• reorder(): Reorders categorical variables based on the values of another variable.