Pair PlotPair Plot

Pair plot is used to plot a pairwise relationship between the numeric variables in a dataset. It is quite similar to a joint plot, however, it is not limited to only two variables. In fact, a pair plot creates an NxN grid of the Axes objects (multiple subplots) where N is the number of numeric variables (numeric columns in a DataFrame).

Let’s have a look at an example of such plot:

Pair plot example

Pair Plot Description

As you can see, for each column x-axis is shared among all the plots in the columns, a certain single variable lies on the x-axis. The same goes for the rows where the y-axis is shared among all the plots in the row. Diagonal plots are histograms by default, since they show the distribution of a single variable (univariate marginal distribution), and the other plots are scatter plots.

Creating a Pair Plot

Creating a pair plot with seaborn comes down to calling its pairplot() function. Its most important and the only required parameter is data which should be a DataFrame object. Here is an example for you:

Here iris_df is the DataFrame we pass in the pairplot() function and everything works just fine. height and aspect parameters just specify the height and width (height * aspect) of each facet (side) in inches.


Another parameter which is worth mentioning is hue which specifies the variable (column name) in data to map plot aspects to different colors or even create separate plots (on one Axes) for each of its values.

Here is an example to make things clear:

You can instantly spot the difference here. First of all, the data points on each scatter plot are colored according to the species they belong to (the respective value in the ’species’ column). Diagonal plots are now KDE plots (a separate one for each of the species) instead of the histograms.

As a matter of fact, when dealing with a classification problem it often makes sense to create a pair plot with the hue parameter set to the target variable (categorical variable we want to predict).

Changing Plot Kinds

You can also set other plots instead of the scatter plots and set other diagonal plots. The parameters kind ('scatter' is its default value) and diag_kind ('auto' is its default value, so its kind is based on the presence of the hue parameter) respectively are used for this purpose.

Let’s now modify our example:

'scatter', 'kde', 'hist', 'reg' are possible values for the kind parameter.

diag_kind can be set to one of the following values:

  • 'auto';
  • 'hist';
  • 'kde'
  • None.

Everything is similar to the jointplot() function in this regard.

More on the pairplot() function in its documentation.


  1. Use the correct function to create a pair plot.
  2. Set the data for the plot to be penguins_df via the first argument.
  3. Set 'sex' as the column which will map the plot aspects to different colors via specifying the second argument.
  4. Set the non-diagonal plots to have a regression line ('reg') via specifying the third argument.
  5. Set height to 2.
  6. Set aspect to 0.8.

It may take a few minutes to verify the solution.

Everything was clear?

Section 5. Chapter 6
toggle bottom row