Exploratory and Explanatory
The terms Exploratory and Explanatory define two distinct phases and purposes of visualization in data science and analysis:
1. Exploratory Visualization (Data Science)
Exploratory Data Analysis (EDA) is doen at the beginning of every Data Science projects. It's done to understand the data. How many categorical features do we have? Are there any imbalances or outliers? It's the first step taken before cleaning up the data. It's basically the process of investigating a dataset to understand its main characteristics, discover patterns, spot anomalies (outliers), and test hypotheses. The visualizations created during this phase are primarily for the analyst themselves.
Purpose
Discovery: To find stories, patterns, and relationships hidden in the data.
Data Quality Check: To identify errors, missing values, or outliers that need cleaning.
Hypothesis Generation: To formulate questions or hypotheses that can be tested later.
Model Preparation: To understand data distributions and feature relationships before building a machine learning model.
Key Characteristics
| Feature | Description |
| Audience | The analyst or data science team. |
| Goal | To understand the data. |
| Design | Messy, unpolished, many charts, fast to generate. |
| Tools | Tools that allow for quick iteration (e.g., Python libraries like Matplotlib, Seaborn, or interactive BI tools). |
| Chart Types | Histograms (for distribution), Box Plots (for outliers), Scatter Plots (for correlation), and Pair Plots (for multivariate relationships). |
2. Explanatory Visualization (Data Analysis)
Explanatory Visualization is the process of presenting results and insights to an audience. This phase happens after the exploratory work is done and a clear message has been established. The goal is to communicate a specific, well-defined story or conclusion.
Purpose
Communication: To convey a specific, pre-determined insight or story.
Decision Support: To guide stakeholders toward a particular conclusion or action.
Persuasion: To present evidence that supports a recommendation or validates a model's findings.
Key Characteristics
| Feature | Description |
| Audience | Stakeholders, managers, clients, or the general public. |
| Goal | To communicate a specific insight and drive action. |
| Design | Highly polished, clean, minimal "chart junk," focuses on one clear message per visual. |
| Tools | Presentation software, reports, or highly curated explanatory dashboards. |
| Chart Types | Bar Charts (for comparison), Line Charts (for trends over time), and highly Annotated Charts (with clear titles and labels that state the conclusion). |
The Analytical Workflow
The two types of visualization are sequential and complementary:
Exploratory Phase: The analyst sifts through 100 different visualizations to find 2 key insights (the "gemstones").
Explanatory Phase: The analyst polishes those 2 key insights into a simple, compelling story for the audience, often using only 2-3 highly targeted visualizations.
In short:
You use Exploratory visualization to ask questions of the data.
You use Explanatory visualization to answer questions for the audience.
Comments
Post a Comment