1. Adding a third variable with hue
We saw in the last lesson that a really nice advantage of Seaborn is that it works well with pandas DataFrames. In this lesson, we'll see another big advantage that Seaborn offers: the ability to quickly add a third variable to your plots by adding color.
2. Tips dataset
To showcase this cool feature in Seaborn, we'll be using Seaborn's built-in tips dataset. You can access it by using the "load dataset" function in Seaborn and passing in the name of the dataset.
These are the first five rows of the tips dataset. This dataset contains one row for each table served at a restaurant and has information about things like the bill amount, how many people were at the table, and when the table was served. Let's explore the relationship between the "total_bill" and "tip" columns using a scatter plot.
3. A basic scatter plot
Here is the code to generate it. The total bill per table (in dollars) is on the x-axis, and the total tip (in dollars) is on the y-axis. We can see from this plot that larger bills are associated with larger tips. What if we want to see which of the data points are smokers versus non-smokers? Seaborn makes this super easy.
4. A scatter plot with hue
You can set the "hue" parameter equal to the DataFrame column name "smoker" and then Seaborn will automatically color each point by whether they are a smoker. Plus, it will add a legend to the plot automatically! If you don't want to use pandas, you can set it equal to a list of values instead of a column name.
5. Setting hue order
Hue also allows you to assert more control over the ordering and coloring of each value. The "hue order" parameter takes in a list of values and will set the order of the values in the plot accordingly. Notice how the legend for smoker now lists "yes" before "no".
6. Specifying hue colors
You can also control the colors assigned to each value using the "palette" parameter. This parameter takes in a dictionary, which is a data structure that has key-value pairs. This dictionary should map the variable values to the colors you want to represent the value. Here, we create a dictionary called "hue colors" that maps the value "Yes" to the color black and the value "No" to the color red. When we set hue equal to "smoker" and the palette parameter equal to this dictionary, we have a scatter plot where smokers are represented with black dots and non-smokers are represented with red dots.
7. Color options
In the last example, we used the words "black" and "red" to define what the hue colors should be. This only works for a small set of color names that are defined by Matplotlib. Here is the list of Matplotlib colors and their names. Note that you can use a single-letter Matplotlib abbreviation instead of the full name. You can also use an HTML color hex code instead of these Matplotlib color names, which allows you to choose any color you want to.
8. Using HTML hex color codes with hue
Here's an example using HTML hex codes. Make sure you put the hex codes in quotes with a pound sign at the beginning.
9. Using hue with count plots
As a final note, hue is available in most of Seaborn's plot types. For example, this count plot shows the number of observations we have for smokers versus non-smokers, and setting "hue" equal to "sex" divides these bars into subgroups of males versus females. From this plot, we can see that males outnumber females among both smokers and non-smokers in this dataset.
10. Let's practice!
We'll be using hue a lot in this course, so let's practice what we've learned to round out the first chapter!