## DEV Community

Muthoni, Rogers

Posted on • Updated on

# Understanding Your Data: The Essentials of Exploratory Data Analysis (EDA)

## Introduction

Exploratory Data Analysis (EDA) is the art of investigating a dataset to discover its main characteristics. It is an important step in every data analysis project. Data professionals often employ EDA techniques to see what the data can reveal while understanding the relationship between different variables in the dataset. Depending on size interpreting and comprehending datasets can be a challenging task. It would not be feasible to make inferences by looking at the first hundred records from a thousand or even a million of them. Through EDA techniques, it becomes easier to extract summaries and find critical and relevant data points that can guide further steps in data analysis projects. In this article, I will discuss various EDA techniques used by data professionals including different methods of summarizing and visualizing data, detecting outliers, and finding correlations.

## EDA Techniques

There are several techniques that form the backbone of EDA. The essential techniques include; summary statistics, analysis of missing data, detection of outliers, correlation analysis, data visualization, time series analysis, exploration of categorical data, and dimensionality reduction. An in-depth discussion of these techniques has been made below.
- Summary Statistics
Summary statistics are useful when you want to get a quick overview of your data. They provide information such as the measures of location and spread. Measures of location (also measures of central tendency) provide information on where the data points are located. The particular measures of location include the mean, median, and mode of the data. Measures of spread tell how data points in the datasets are varied or spread out. The particular measures of spread include quartiles, range, interquartile range, variance, and standard deviation.

_- Analysis of Missing Data _
Missing data refers to values or data points that may not be present for some columns (or features or variables) in a given dataset. This may be caused by different factors such as human error or faulty sensors used during data collection. The need to address missing data in data analysis is critical for the success of the project and the reliability of the resulting models.
Missing data may be represented in different forms. Common representations of missing data include; NaN, NULL, or blank spaces. Understanding how missing data is represented in a dataset is important as it will determine the best approach for data cleaning.

_- Detection of Outliers _
During EDA, data professionals often encounter outliers in the data. Outliers are data points that significantly deviate from the general behavior of the dataset. In practice, these data points tend to lie far away from the rest of the data. Outliers may be contributed by errors or data corruption. Although this is the case, outliers could also represent genuine extreme values in the dataset.
Outliers in a dataset can affect statistical analysis, machine learning models, and results of data visualization techniques. Therefore, these values must be addressed properly to avoid producing biased results and conclusions.
Different techniques can be employed to identify and remove outliers. Common techniques include the use of visualizations (such as boxplots) and statistical methods (such as percentiles and z-scores)

- Correlation Analysis
In EDA, correlational analysis is used to show the degree of association between different features in a dataset. Such information can be used during subsequent steps in data analysis especially when there is a need to understand how different variables in the dataset are related. Correlation analysis can also indicate different factors confounding the relationship aspect under study.

_- Data Visualization _
Data Visualization is an important technique in EDA. It allows data professionals to “look at” the data and have a glimpse of how variables are related to each other. This involves the use of graphs and charts that make it easier to spot patterns, trends, anomalies, and relationships. Histograms, bar plots, line plots, and scatter plots are common visualizations that are used in visualizing how variables are related.

• Time Series Analysis
• Categorical Data Analysis
• Dimensionality Reduction