2.6: Exploring geom Functions

Plotting in ggplot2

Data Visualization Types and When to Use Them

Scavenger Hunt

Head over to data-to-viz.com and do some exploration! In your groups, choose one of the following types of plots. Hunt down as many locations in the flow chart where your plot type is found as you can find.

  • histogram: geom_histogram()
  • scatter plot: geom_point()
  • box-and-whisker: geom_boxplot() + geom_jitter()

If you’ve finished, click on your plot type and read more about it. Check out the examples that were made in R, including many using ggplot2. Take note of the following:

  • which geom function is being used to create the plot
  • you’ll see a number of other things in the ggplot code, some which will be new. What are they? Any idea what they are doing?

Using geom Functions to Make Plots

Let’s practice making different kinds of plots with various geom functions to see how they work.

library(tidyverse)
fish <- read_csv("data/fish_sick_data.csv")

Plotting in ggplot2

First, let’s remind ourselves of the general structure of how we make plots using the ggplot2 syntax.

# ggplot(data = <DATA>, mapping = aes(x = <COLUMN1>, y = <COLUMN2)) +
# geom_function() +
# labs() +
# theme()

Histograms

As we learned last modules, histograms are plots which let us look at one continuous variable.

They help us get a feel for the distribution of that data. To make histograms in ggplot2, we use the geom_histogram() function. Let’s look at the number of fish.

ggplot(fish, aes(num_fish)) +
  geom_histogram(bins = 20)

# we can change the number of bins (essentially, the number of columns) by modifying the bins argument in the geom_histogram function

Let’s Practice

Make a histogram of the number of sick fish in the tanks. Create the histogram with 10 bins (10 groupings).

ggplot(fish, aes(num_sick)) +
  geom_histogram(bins = 10)

Multiple Histograms (with geom_histogram())

When we create a multiple histogram, we have to add one additional argument to geom_histogram(). Let’s see what happens if we just specify fill.

ggplot(fish, aes(num_fish, fill = species)) +
  geom_histogram(bins = 10)

What is happening in the column with both green and red? Perhaps the teal histogram is in front of the red histogram and as blocking us from seeing some red?

Let’s change the transparency using an argument called alpha, which allows us to make layers transparent. The scale for alpha goes from 0 (completely transparent) to 1 (not transparent at all).

We can set the transparency to 0.5 to see if there is any overlap.

ggplot(fish, aes(num_fish, fill = species)) +
  geom_histogram(bins = 10, alpha = 0.5)

Not too much changed. It still doesn’t look like we can see any red points behind the teal. Perhaps the teal values and red values are stacked on top of one another?

Let’s take a look at what happens when we add the argument position = "identity".

ggplot(fish, aes(num_fish, fill = species)) +
  geom_histogram(bins = 10, alpha = 0.5, position = "identity")

Aha! This is different from above. Instead of red being stacked vertically on top of teal, we can now see that the red values start at 0 on the y-axis and are overlapping with the teal.

The position = "identity" argument tell geom_histogram to plot the data for each group starting from 0 in the y-axis rather than stacking values from the same group, which is the default.

Any time we plot a multiple histogram, we need to change the transparency and add position = "identity" when creating a multiple histogram; if we don’t, we won’t see potential overlap!

Let’s Practice!

Make a multiple histogram of the number of sick fish per species. Make sure your plot has 10 bins, is partially transparent, and the data are not stacked.

ggplot(fish, aes(num_sick, fill = species)) +
  geom_histogram(bins = 10, alpha = 0.5, position = "identity")

The powerful and pesky aes() function

A quick note about the aes() function. It’s one of the more confusing bits of ggplot2.

When do I put the color (or size or linetype or fill or whatever) inside the aes() function versus in the geom function but outside of aes()? When we made our density plots, why did the color argument go inside of aes() but alpha went outside?

Essentially, it boils down to this:

  • if you want something (color, size, etc.) on your plot to change based on a variable from a data frame, you will want to put the argument within the aes() function.

  • if you want something (color, size, etc.) on the plot to be constant, you will specify it outside of the function.

For some additional examples and explanation, check out this Stack Overflow page.

Scatter Plot

As a reminder, we use the geom_point() function to make a scatter plot of the relationship between two continuous variables.

ggplot(fish, aes(avg_daily_temp, num_fish)) +
  geom_point()

Let’s Practice: Multiple Scatterplot

Using what you’ve learned about making histograms, see if you can create a “multiple scatterplot,” where the color of the points are determined by the fish species.

Hint: you’ll want to use an argument called color.

ggplot(fish, aes(avg_daily_temp, num_fish, color = species)) +
  geom_point()

Box-and-Whisker Plots

Box-and whisker-plots (also known as box plots) are another great option for looking at one continuous variable and one or more categorical variables. They are particularly nice when you want to see measures of central tendency and variation in the same plot.

Let’s build one and then talk through what each component means. We use geom_boxplot to make these types of plots.

ggplot(fish, aes(species, num_fish, color = species)) +
  geom_boxplot()

So what does the box represent? And the whiskers?

  • the box represents the middle 50% of the values in the data set.
  • the line that runs through the middle of the box represents the median (middle value) of the data
  • the whiskers represent the spread of the data (we won’t get into the mathematical details of exactly how they are calculated) and are roughly comparable to 95% confidence intervals (we will cover this in another module)
  • values that fall outside of the whiskers can be considered outliers and are plotted individually

Layering

One of the beautiful parts of working with ggplot2 is that you can add multiple layers to each plot.

One of the key things missing from box-and-whisker plots is any indication of how many data points we have. In the plot above, there could be 5 tanks per species or 500 tanks per species.

How can we add an indication of how many points there are? We can layer each individual data point on top of the boxes!

ggplot(fish, aes(species, num_fish, color = species)) +
  geom_boxplot() +
  geom_point(alpha = 0.5)

This is nice, but there is still some overlap in points that makes it hard for us to see exactly how many points are there.

The geom_jitter() function is a special version of geom_point(). It adds a little bit of randomness to the points (both horizontally and vertically) so that they don’t overlap as much.

We can control how much randomness we allow with the width and/or height arguments. I usually ignore height and set the width argument to 0.1.

ggplot(fish, aes(species, num_fish, color = species)) +
  geom_boxplot() +
  geom_jitter(alpha = 0.5, width = 0.1)

That is looking really nice! We can keep improving it, though, with better labels for the axes and the legend as well as a nice theme.

ggplot(fish, aes(species, num_fish, color = species)) +
  geom_boxplot() +
  geom_jitter(alpha = 0.5, width = 0.1) +
  labs(x = "Species",
       y = "Number of Fish Per Tank",
       color = "Species") +
  theme_light()

color and fill

We have used two different arguments in the aes() function to specify that we want to colors in our plots to change based on one of the columns in our dataset. When do you want to use which one?

color

For the most part, the color argument refers to the color of the lines in a plot (e.g., in box plots, the colors of the lines change).

The one exception to this is in geom_point(). We change the color of the points with the color argument, as well.

fill

When there is space that we want to fill in with color depending on the values in a column, we want to use the fill argument.

For histograms, the fill argument changes the color inside the bins. If we use color in a histogram instead, only the outlines of the columns will change, but the columns will remain filled with gray.

In boxplots, the inside area of the box will be filled with color, but the lines will stay black.

In Labels

If we have specified either color or fill (or both) in the aes() function, ggplot will automatically create a legend for us. The key is determined by whichever argument we used.

So, if you use the color argument in the aes() function, you would want to use the color argument in the labs() function to change the title of the legend. If you used the fill argument, you would then use the fill argument in the labs() function.

If you have both color and fill in aes(), you will need to add both arguments to the labs() function. If you give the same label to each, it will create one key.

Let’s Practice!

Make 2 different types of plots with the same data: the average daily temperature of the tanks and fish species.

First, think about the variables we are plotting. How many are there? Are they qualitative or quantitative?

Based on those answers, determine which plot types you can produce for those variable types.

Now, make your two plot! To each, add labels and a theme.

# Boxplot
ggplot(fish, aes(species, avg_daily_temp, color = species)) +
  geom_boxplot() +
  geom_jitter(alpha = 0.5, width = 0.1) +
  labs(x = "Average Daily Temp (C)",
       y = "Frequency", 
       color = "Fish Species") +
  theme_bw()

# Histogram
ggplot(fish, aes(avg_daily_temp, fill = species)) +
  geom_histogram(bins = 10, alpha = 0.5, position = "identity") + 
  labs(x = "Average Daily Temp (C)", 
       y = "Frequency",
       fill = "Fish Species") +
  theme_bw()

Finished? Let’s add another variable. Now, we want to plot average daily temperatures, the number of sick fish, and the fish species. Work through the same steps as above.

ggplot(fish, aes(avg_daily_temp, num_sick, color = species)) +
  geom_point(alpha = 0.5) +
  labs(x = "Average Daily Temp (C)",
       y = "Number of Sick Fish per Tank",
       color = "Fish Species") +
  theme_light()

Summary: Data Visualization Types and When to Use Them

Let’s summarize some of what we’ve learned in this module

Histogram or Density Plot

Good for looking at the distribution of one continuous variable

  • one continuous variable (x-axis)

Multiple Histogram or Multiple Density Plot

Good for looking at differences in the distributions of one continuous variable based on a categorical variable

  • one continuous variable (x-axis)
  • one categorical variable via the fill or color argument in the aes() function
  • we always want to add transparency (alpha) and, for histograms, position = "identity"

Scatter Plot

Good for looking for the relationship between two continuous variables

  • two continuous variables (x-axis and y-axis)
  • can add in a categorical variable via aes(), but the main relationship is between the two continuous variables

Box Plot

Good for looking at measures of central tendency and variation for a continuous variable and the differences between those measures between categories

  • one continuous variable (y-axis)
  • at least one categorical (x-axis and additional via aes())

Layering

We can add multiple layers to ggplots, which is part of what makes them so useful!

  • we can add multiple geom functions to a single plot
  • we use the labs() function to rename axes labels and legends
  • we use a theme function to make the plot more aesthetically pleasing and easier to understand