5 Best Ways to Draw a Boxplot for Each Numeric Variable in a DataFrame with Seaborn

by

💡 Problem Formulation: When exploring data, visualizing the distribution of numeric variables is invaluable. Data scientists often want to draw boxplots for each numeric variable in a pandas DataFrame using Seaborn, which is a powerful visualization library in Python. Assume we have a DataFrame with multiple numeric columns, and we want to quickly generate boxplots for an at-a-glance comparison of their distributions.

Method 1: Standard Boxplot with seaborn.boxplot()

This method employs Seaborn’s boxplot() feature, which is straightforward and allows detailed customization of the resulting boxplots. You can specify which DataFrame columns to plot, as well as the orientation of the boxplots.

Here’s an example:

import seaborn as sns
import pandas as pd

# Creating a sample DataFrame
data = pd.DataFrame({
    'Variable1': [10, 20, 30, 40],
    'Variable2': [5, 15, 25, 35],
    'NonNumeric': ['A', 'B', 'C', 'D']  # Non-numeric column to be ignored
})

# Select only numeric columns for boxplot
numeric_data = data.select_dtypes(include=['number'])

# Drawing the boxplot
sns.boxplot(data=numeric_data)

The output is a series of boxplots, each representing the distribution of one of the numeric columns in the DataFrame.

This snippet filters out non-numeric columns using pandas.DataFrame.select_dtypes() and then uses seaborn.boxplot() to plot boxplots for the remaining columns. It is an explicit approach that provides clarity on the data being plotted.

Method 2: Boxplot with Automatic Numeric Column Detection

Seaborn can automatically detect numeric columns when using boxplot(). This method is convenient when you have a DataFrame with both numeric and non-numeric columns and you wish to plot only the numeric ones without additional filtering.

Here’s an example:

import seaborn as sns
import pandas as pd

# Create a mixed-type DataFrame
data = pd.DataFrame({
    'Score': [88, 92, 85, 99],
    'Age': [20, 23, 22, 21],
    'Gender': ['F', 'M', 'M', 'F']
})

# Directly plotting the boxplot, Seaborn automatically selects numeric columns
sns.boxplot(data=data)

The output will display boxplots for ‘Score’ and ‘Age’ columns, automatically excluding the non-numeric ‘Gender’ column.

Here, seaborn.boxplot() is clever enough to ignore non-numeric data, making it a quick and efficient method to visualize numeric distributions without preprocessing.

Method 3: Faceted Boxplot with seaborn.catplot()

seaborn.catplot() can create faceted boxplots, which means plotting multiple boxplots across a grid layout. This function is useful for comparing distributions across categories or over a particular subset.

Here’s an example:

import seaborn as sns
import pandas as pd

# Creating a DataFrame with an additional categorical column for faceting
data = pd.DataFrame({
    'Exam1': [90, 80, 85, 88],
    'Exam2': [78, 82, 89, 94],
    'Class': ['A', 'B', 'A', 'B']
})

# Using catplot to create a faceted boxplot layout
sns.catplot(kind='box', data=data, col='Class')

The output is two boxplots for each ‘Exam’ per class category, laying side by side for easy comparison.

The code snippet utilizes seaborn.catplot() to create a gridded boxplot visualization. By setting the kind parameter to ‘box’ and specifying a categorical column for col, we can compare the numeric data distributions across different categories.

Method 4: Pairwise Boxplots with seaborn.pairplot()

The seaborn.pairplot() function is typically used for pair-wise relationships in a dataset, but it can also be configured to display boxplots along the diagonal. This method is best for simultaneously exploring correlations and distributions.

Here’s an example:

import seaborn as sns
import pandas as pd

# Creating a DataFrame with numeric variables
data = pd.DataFrame({
    'Test1': [32, 45, 50, 39],
    'Test2': [22, 39, 40, 29],
    'Test3': [12, 35, 30, 19]
})

# Creating pairplot with boxplots on the diagonal
sns.pairplot(data, diag_kind='box')

The output is a grid of scatter plots for each variable pair and boxplots along the diagonal showing distributions for each individual variable.

This code uses seaborn.pairplot() to create a matrix of scatter plots and specifies that boxplots should be used on the diagonal with the diag_kind='box' argument. This method gives a comprehensive overview of the relationships and distributions among multiple numeric variables.

Bonus One-Liner Method 5: Boxplot with Pandas Plotting Backend

In Pandas versions 0.25 and later, we can leverage the built-in plotting backend for a quick one-liner boxplot. Here we tell pandas to use Seaborn as its backend for plotting.

Here’s an example:

import pandas as pd
pd.options.plotting.backend = "seaborn"

# Creating a simple DataFrame
data = pd.DataFrame({
    'Feature1': [1, 2, 3, 4],
    'Feature2': [4, 3, 2, 1]
})

# One-liner to draw the boxplot
data.plot(kind='box')

The output displays a boxplot for each feature in the DataFrame, using Seaborn’s style.

This code snippet changes the pandas plotting backend to Seaborn with pd.options.plotting.backend = "seaborn", then uses the simple data.plot(kind='box') command to draw the boxplots. It’s the quickest method for those running Pandas 0.25+.

Summary/Discussion

  • Method 1: seaborn.boxplot(). Comprehensive and customizable. Requires explicit handling of non-numeric columns.
  • Method 2: Automatic Numeric Detection. Quick, with no need for manual filtering, but less explicit regarding the columns being plotted.
  • Method 3: seaborn.catplot(). Ideal for faceted boxplots, but involves slightly more complex syntax and handling of categorical variables.
  • Method 4: seaborn.pairplot(). Offers a combination of scatter plots and boxplots for examining correlations and distributions. More informative but potentially overwhelming with too much data.
  • Method 5: Pandas Plotting Backend. Easiest and fastest for those using recent versions of Pandas. Limited customization compared to native Seaborn methods.