Using Seaborn’s FacetGrid Based Methods for Exploratory Data Analysis

Introduction

Exploratory Data Analysis or EDA is a vital step in the Data Science Life Cycle. A lot of primary observations and insights are drawn during the Exploratory Data Analysis step. Apart from this, we also perform hypothesis testing, outlier analysis and necessary data cleaning. In all these processes, data visualization plays an imperative role. For visualizing, we often use charts and plots for Uni-variate and Bi-variate Analysis such as Count Plots, Box Plots, Histogram, Scatter Plots et Cetra.

We can plot these charts for each of the features in our dataset but this can be time-consuming when we have limited time to present our insights. Of course, we can define a function to do the tasks in lesser lines of code, but when we talk about drawing insights quickly from data, we need something that can accomplish the task in a single command. Luckily, we have few methods with us to do the same.

In this article, we will explore some methods that are very useful in drawing insights in no time from our data. Although, it gives limited information, but drawing primary observations from our data, when e don’t have enough time in our hands, is a boon.

FacetGrid and the EDA

Exploratory Data Analysis helps in identifying the underlying patterns and trends which are not observable just by looking at the data. For example, plotting a distribution plot for a numeric feature, say Age in a dataset displays the distribution of Age of the people for the continuous interval. A scatter plot of two variables would help in determining the relationship between the variables how the increase or decrease in one affects the other.

To accomplish the task, our fundamental libraries Matplotlib and Seaborn have all the required methods. We just have to pass the dataset we are working on as an argument, specify the column we want to explore and we will get our plots within seconds. The methods we are going to use are will plot on Seaborn’s FaceGrid. A FacetGrid is a multi-axes grid with subplots visualizing the distribution of variables of a dataset and the relationship between multiple variables. Now, let’s see the methods which can help to perform EDA in no time.

The .catplot() Method

Seaborn’s .catplot() method plots the relationship between numerical and one or more categorical variables onto the FacetGrid. Seaborn library is built over the Matplotlib thus it has all the functionalities of including the recovered drawbacks of Matplotlib. For the proper functionality and usability of catplot, we need to pass 3 main parameters, i.e. the data, variable to be plotted on the x-axis and variable to be plotted on the y-axis. To increase usability, we often pass ‘col’ and ‘hue’ parameters which also takes features from our dataset, preferably categorical as that’s what our motive is.

Let’s understand this with an example.

Importing the Libraries

import seaborn as sns

Importing the data

df = sns.load_dataset('mpg')

Here, we are using the ‘mpg’ dataset from Seaborn’s data library.

Checking the Dataset

df.head()

Plotting the catplot

Here we decided to plot the number of cylinders by the displacement of the vehicle.

sns.catplot(data=df, x="cylinders", y="displacement");

Putting it all together
Python Code:

import seaborn as sns
import matplotlib.pyplot as plt
df = sns.load_dataset('mpg')
# df.head()
sns.catplot(data=df, x="cylinders", y="displacement")
plt.show()

On executing this code, we get:

 

Increasing Usability of catplot

To make the most out of the catplot, we will define additional parameters which will help us explore the data to a larger extent and drawing insights.

1. Specifying the ‘col’ Parameter

Specifying the col parameter will determine the faceting of the plot. For example, if we specify a categorical feature into the ‘col‘ parameter which has 3 unique values, we will have 3 different plots, each for each unique value of the categorical variable used in the FacetGrid. Let’s understand it with an example:

import seaborn as sns
df = sns.load_dataset('mpg')
# df.head()
sns.catplot(data=df, x="cylinders", y="displacement", col = 'origin');

On executing this code, we get:

Image Source – Personal Computer

In this, we can see, we get 3 different plots for 3 kinds of origin in our dataset, namely ‘usa‘, ‘japan‘, and ‘europe‘.

2. Specifying the ‘hue’ Parameter

Specifying the ‘hue‘ parameter will specify the colour of the data points based on the feature used in the ‘hue’ parameter. For example,

import seaborn as sns
df = sns.load_dataset('mpg')
# df.head()
sns.catplot(data=df, x="cylinders", y="displacement", col = 'origin', hue = 'model_year');

On executing this code, we get:

Image Source – Personal Computer