Data analysis

Why did revenue fall in September? Which customers stop buying? Is the new cart page better than the old one? This course is about how you answer a question like that with data: how you get it and clean it, how you see what is in it, and how you decide honestly what follows from it.

The code is Python, with the libraries every analyst uses: NumPy, pandas, matplotlib and scikit-learn. It runs here, in the browser: every code field is real Python. You write, press “Run” and see the table or the chart under the code. Nothing to install.

The first run downloads Python (about 13 MB), and pandas with the charts about 17 MB more. The browser keeps them, so afterwards everything starts at once. Machine learning at the end of the course adds about 19 MB.

How you learn here

The course is split into modules, and every module into short lessons. A lesson is a series of cards, one idea per card: a question from the shop, the code that answers it, and a small task you check straight away. A Python error is explained in plain words, with the original underneath.

The data is the same shop as in the SQL course: products, customers, orders and order lines, in euros. All of it is made up, and the numbers are the same on every run.

You need Python up to collections and functions (lists, dicts, for, functions). SQL helps, but you do not need it. Every module uses only what the earlier ones taught, and your progress is kept in the browser.

The course map

Module 0

Welcome

2 lessons · what analysis does, the sandbox, the first table

1. 1Welcome
2. 2The first table

Module 1

NumPy

4 lessons · arrays, maths without for, summaries, broadcasting

1. 1The NumPy array
2. 2Vectorization
3. 3Summaries and missing values
4. 4Broadcasting: the number, the row and the column

Module 2

Tables with pandas

5 lessons · Series and DataFrame, selecting, filters, new columns, types and gaps

1. 1Series and DataFrame
2. 2Selecting: [], loc and iloc
3. 3Filters and sorting
4. 4New columns
5. 5Types and missing values

Module 3

Wrangling data

5 lessons · groups, joins, wide and long, dates, dirty data

1. 1Groups: groupby
2. 2Joining: merge
3. 3Wide and long: pivot_table and melt
4. 4Dates and time series
5. 5Data cleaning

Module 4

Charts

4 lessons · matplotlib, which chart for which question, distributions, honest charts

1. 1The first chart
2. 2Which chart for which question
3. 3Distributions
4. 4Honest charts

Module 5

Statistics for decisions

6 lessons · noise, intervals, hypothesis tests, an A/B test, correlation and cause

1. 1Describing the data
2. 2Sample and population
3. 3Confidence intervals
4. 4Testing a hypothesis
5. 5A/B test
6. 6Correlation and cause

Module 6

Models with scikit-learn

5 lessons · prediction, classification, leakage, an honest check

1. 1Models: learning from examples
2. 2Linear regression
3. 3Classification
4. 4Data leakage
5. 5Cross-validation

Module 7

Project: why revenue fell

1 lesson · the whole path with the shop data

1. 1Why did revenue fall?

Practice

Tasks

Every task from the lessons in one place, by module.