Data science is the practice of learning from data. In simple terms, it means using information to answer questions and make better decisions. It brings together three pillars—computation, statistics, and domain knowledge—to discover patterns and evidence that help us understand, predict, and improve the world around us. You have already seen data science in action many times without noticing it.
This chapter introduces what data science is and what it is not, setting the stage for the hands-on tools you will learn in Part I.
Data are pieces of recorded information.
They can be:
Some data are structured, meaning they are organized in a clear and consistent format. Often this format is a table with rows and columns, like a spreadsheet or database table. Structured data are easy for computers to sort, filter, and analyze.
Other data are unstructured, meaning they do not come neatly organized. A long paragraph of text, a photograph, or an audio recording are all examples of unstructured data. They still contain information—but that information may need to be transformed into a more organized format before certain types of analysis are possible.
It is important to note that data do not have to fit into rows and columns to be real data. Images, videos, and text can be analyzed directly using specialized tools. However, many common statistical methods work most easily when data are arranged in a structured format.
To see this more clearly, imagine a small table about students in a class:
| Name | Grade | Favorite_Subject |
|---|---|---|
| Alex | 88 | Math |
| Jordan | 92 | History |
| Sam | 75 | Biology |
Each row is an observation (one student).
Each column is a variable (a recorded characteristic).
When information is organized like this, a computer can sort it, analyze it, and visualize it efficiently.
Here is that same table created in Python.
import pandas as pd
# create a small structured dataset
students = pd.DataFrame({
"Name": ["Alex", "Jordan", "Sam"],
"Grade": [88, 92, 75],
"Favorite_Subject": ["Math", "History", "Biology"]
})
students| Name | Grade | Favorite_Subject | |
|---|---|---|---|
| 0 | Alex | 88 | Math |
| 1 | Jordan | 92 | History |
| 2 | Sam | 75 | Biology |
Now the information is structured data that a computer can analyze directly.
If instead we had written:
“Alex scored 88 and likes Math. Jordan scored 92 and likes History…”
that would still be data—but it would not be organized in a consistent structure. Before we could apply many standard statistical tools, we would likely reorganize it into a table.
Can you think of an example of data that would be difficult to fit into rows and columns?
Every data science project begins with a question.
Why has traffic increased on my school’s street?
Do certain neighborhoods file more noise complaints?
Which players contributed most to a team’s winning season?
But not every question works well for data science. Good data science questions are specific and measurable. That means we can connect them directly to data.
For example:
Vague question:
Why are students stressed?
More precise question:
How many hours do students report studying per week, and how does that relate to reported stress levels?
The second question can be answered with collected data. The first one is too broad without further definition.
We can even simulate small data to make the question concrete.
import numpy as np
np.random.seed(42)
hours_studying = np.random.randint(1, 6, size=10)
stress_level = np.random.randint(1, 6, size=10)
list(zip(hours_studying, stress_level))[(np.int64(4), np.int64(4)),
(np.int64(5), np.int64(3)),
(np.int64(3), np.int64(5)),
(np.int64(5), np.int64(2)),
(np.int64(5), np.int64(4)),
(np.int64(2), np.int64(2)),
(np.int64(3), np.int64(4)),
(np.int64(3), np.int64(5)),
(np.int64(3), np.int64(1)),
(np.int64(5), np.int64(4))]This simple example generates pairs of numbers: (hours studying, stress level). Now we could measure whether higher study time is associated with higher stress.
Data science helps translate everyday curiosity into clear, answerable questions. Once a question becomes measurable, we can analyze it, test ideas, and draw conclusions based on evidence.
Can you think of a question about your school that could be rewritten to make it measurable?
Data science is not only about ideas. It is also about building systems that allow you to work carefully and efficiently.
While mathematics and statistics help us reason about data, data scientists spend much of their time using computational tools. These tools allow us to load data, clean it, analyze it, and communicate results in a clear and repeatable way.
Some of the most important tools you will learn include:
Each tool has a specific role.
The command line allows you to navigate folders, create files, and run programs without clicking through menus. This makes your work faster and easier to automate.
Git records every change you make to a project. If you make a mistake, you can return to an earlier version. If you collaborate with others, Git keeps everyone’s work organized.
An IDE (Integrated Development Environment) is more than a simple text editor. It allows you to write code, run programs, manage files, and debug errors in one place. Many different IDEs exist, including free and open-source options. The important idea is not the brand of software, but the workflow it supports.
Quarto allows you to write explanations and code in the same document. When you render the file, Quarto runs the code and inserts the results directly into your report. This connects your reasoning and your computation in one reproducible workflow.
You may already be familiar with graphical user interfaces (GUIs), where you click buttons and select options from menus. GUIs are useful for quick exploration and one-time tasks. However, steps performed only by clicking are often difficult to document precisely.
For this reason, professional data scientists prefer writing code. Code records every instruction explicitly. That makes analyses easier to verify, repeat, and improve.
For example, here is a small calculation written in Python:
numbers = [3, 7, 10]
sum(numbers)20Instead of calculating by hand or clicking through menus, you write the instruction once and let the computer execute it exactly.
These tools work together. The command line manages your environment. Git tracks your changes. Your IDE organizes and runs your code. Quarto publishes your results. Python performs the computation.
Learning these tools gives you the ability to actually do data science, not just talk about it.
Graphical user interfaces (GUIs) allow you to click buttons and choose options from menus. They are helpful for quick exploration and small tasks.
However, when you perform steps only by clicking, those steps may not be recorded in a clear, repeatable way. If someone asks, “Exactly what did you do?”, it can be difficult to answer precisely.
Code works differently. Every instruction is written down. That means the computer — and other people — can rerun the exact same steps.
In data science, we prefer tools that make our work transparent and reproducible.
Why might writing code be more reliable than performing calculations by hand or clicking through menus, especially for large datasets?
Reproducibility is a core principle of data science. It means that someone else, or even you in the future, can run the same analysis and obtain the same results.
In professional settings, results must be verifiable. If a report claims that average test scores increased by 5 points, others should be able to examine the data and confirm the calculation. Without reproducibility, findings are difficult to trust.
Imagine finishing a project and coming back to it six months later. Would you remember:
If the answer is no, the work is not reproducible.
A reproducible workflow is one where:
For example, suppose you calculate the average test score.
scores = [88, 92, 75]
sum(scores) / len(scores)85.0If that calculation lives inside your script, anyone can rerun it. But if you computed it on a calculator and only wrote down the final number, no one can verify how you obtained it.
Reproducibility protects you from mistakes. If something looks wrong, you can rerun everything from the beginning. It also builds trust. Others can inspect your code and confirm your results.
This is why data scientists rely on tools such as:
Together, these tools allow an entire project to be rebuilt from scratch using documented steps.
If a classmate wanted to reproduce your project, what files would they need? Would a screenshot of your results be enough?
Data science is interdisciplinary. That means it combines knowledge from multiple fields rather than relying on just one.
To answer real-world questions, data scientists bring together three main pillars:
Each pillar plays a different role.
Computing allows us to handle large datasets efficiently. Without code, it would be nearly impossible to analyze millions of records.
Statistics helps us reason carefully about patterns. It teaches us how to distinguish real signals from random variation. Without statistics, we might mistake coincidence for evidence.
Domain knowledge provides context. Data do not explain themselves. If you are analyzing hospital data, you need basic knowledge of health care systems. If you are studying sports performance, you need to understand the rules of the game.
Consider an example. Suppose a city wants to predict flooding.
If any one pillar is missing, the analysis becomes weaker.
Computing without statistics may produce precise but misleading numbers. Statistics without computing may be too slow to apply at scale. Both without domain knowledge may lead to unrealistic conclusions.
Data science works best when these three areas support each other.
Think of a problem you care about (sports, music, climate, school policy). What kind of domain knowledge would you need to analyze it properly?
Data science is not a shortcut around mathematics and statistics. When students try to rely only on software without understanding probability, variation, or logical reasoning, they often produce results that look polished but are unreliable. A model can generate precise numbers, yet those numbers may rest on weak assumptions. Strong quantitative training does not make work complicated for its own sake; it prevents false confidence.
Data science is also not an exercise in writing code for its own sake. A well-written program can process enormous amounts of data, but speed does not guarantee truth. If the underlying logic is flawed, computing simply amplifies the mistake. Learning to program well means learning to think carefully about algorithms, structure, and consequences.
Most importantly, data science is not magic. No method can extract reliable information from pure noise. Humans are naturally drawn to patterns, even when they occur by chance. Without statistical discipline, it is easy to mistake randomness for discovery. History is full of examples where weak evidence led to strong claims, causing confusion or harm. Responsible data science requires skepticism, patience, and respect for limits.
Part I of this book focuses on building your foundation:
By the end of Part I, you will have built your first complete data science project: a real report, with real data, that you can publish or share.
Part II takes you through case studies in health, environment, sports, and other areas—helping you discover patterns in real datasets and understand how data science works in practice.
Data science is not something you learn in one day. It is a craft you build over time. This book gives you the tools and the mindset to begin that journey the right way.
These questions are designed to help you think more carefully about the ideas in this chapter. Write short answers unless instructed otherwise.
In your own words, define data science. What are its three main pillars?
Give one example of structured data and one example of unstructured data from your daily life.
Rewrite the following vague question so that it becomes measurable:
“Are students happy at our school?”
Explain why clicking through menus in a GUI might make an analysis harder to reproduce.
Imagine you conducted a small survey of 10 classmates about how many hours they sleep each night.
Why is reproducibility important in professional settings? Give one practical reason.
Choose a real-world topic you care about (sports, music, climate, social media, school policy, etc.).
Optional CLI Warm-Up.
Open your terminal.
Type:
pwdWhat does this command display?
Now type:
lsWhat do you see?
In one sentence, explain how using the command line differs from clicking through folders.
Mini-Project Preview.