Applied Statistics in Python, One Proof at a Time
Applied Statistics in Python, One Proof at a Time
Free first edition / MVP edition
Version: v0.1
Release date: 2026-06-12
Publisher: NEKpress.ca
This compact first edition introduces PyStatsV1 as an MIT-licensed educational statistics project. It is designed to be downloaded, read, and used as a practical starting point before a learner commits to a longer course or textbook.
Reader promise: by the end of this short book, you should understand what PyStatsV1 is for, how to prepare a safe Python environment, how to run a small first proof, and how to read early output without overclaiming.
The goal is simple: install the tools, run a small proof, inspect the output, and learn how to make statistical work more reproducible.
PyStatsV1 is an MIT-licensed applied statistics learning project built around Python, reproducible examples, and inspectable outputs. It is not trying to hide statistics behind a button. It is trying to help a learner see the command, the data path, the output path, and the check that proves the result was produced.
The project is educational. It is useful for learning, practice, demonstrations, and small proof-first workflows. It is not a substitute for professional review when real decisions depend on the data.
A good PyStatsV1 learning loop has four parts:
This is the same proof-first habit used throughout the LearnToProgram.ca portal. You do not just hope the tool works. You ask the tool to leave evidence.
This book is for careful beginners, students, tutors, instructors, analysts, and self-directed learners who want a practical way to learn applied statistics in Python. It assumes you can use a terminal at a basic level, but it does not assume that you are already an expert programmer.
This book is not statistical consulting, legal advice, medical advice, financial advice, privacy advice, tax advice, or payment-compliance advice. It is an educational starting point. For real research or business decisions, review assumptions, data quality, and the consequences of error with appropriate professional care.
Before installing any statistics package, start by checking your Python environment. A reproducible workflow begins with a known interpreter, a known project folder, and a small set of commands that can be repeated later.
Use a project folder you control. Avoid installing experimental learning tools into a system Python environment. A virtual environment keeps the learning project separate from the rest of your computer.
A common Ubuntu or macOS workflow is:
python3 --version
python3 -m venv .venv
source .venv/bin/activate
python -m pip install -U pipOn Windows, use PowerShell and a project folder you control:
py -3 --version
py -3 -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -U pipThe learner-facing path is expected to support a workbook extra:
python -m pip install "pystatsv1[workbook]"
pystatsv1 doctorThe source-inspection path is useful when you want to read the code and examples directly:
git clone https://github.com/pystatsv1/PyStatsV1.git
cd PyStatsV1
python -m venv .venv
source .venv/bin/activate
python -m pip install -U pip
python -m pip install -e .The first proof is not a statistical result. The first proof is that your environment can run the project and report its status.
pystatsv1 doctorSave the command and the output. That small habit helps later when an installation works on one machine but fails on another.
A workbook is useful because it gives the learner a stable place to run scripts, inspect outputs, and compare results. The workflow should be boring in the best possible way: initialize, run, inspect, test, explain.
A typical first pass is:
pystatsv1 workbook init
cd pystatsv1_workbook
pytest -qIf tests pass, the workbook is not just present. It has produced evidence that the baseline examples are runnable in your environment.
After the workbook exists, pick one small script. Avoid jumping straight to a complex model. A useful first example is a descriptive-statistics or introductory hypothesis-testing script.
python scripts/psych_ch10_problem_set.py
pytest -qRead the output before changing anything. Ask three questions:
A statistic is not a conclusion by itself. A p-value, confidence interval, regression coefficient, or diagnostic plot needs context. Before writing a claim, identify the question, the data source, the model or test, and the assumptions that could matter.
A safer sentence often looks like this:
In this worked example, the script produced a difference consistent with the chapter exercise. The result is evidence for the exercise workflow, not a general claim about a population outside the example.That sentence is modest on purpose. PyStatsV1 is meant to help you build reliable habits before you make broad claims.
A two-sample comparison is a good teaching example because it forces the learner to separate the question from the calculation. The statistical tool can compare groups, but the learner must still ask what the groups mean and whether the comparison is fair.
A proof-first mini workflow is:
python scripts/psych_ch10_problem_set.py
pytest -qThen write a short review note:
The point is not to memorize one command. The point is to learn how a command, an output, and a test fit together.
Reliability is about whether a measurement process is stable enough to be useful. In statistics education, reliability examples help learners see that data quality comes before flashy modeling.
A proof-first reliability note should include:
If the result is based on synthetic or classroom data, say so. Synthetic data is valuable for learning because it is safe and reproducible, but it should not be presented as field evidence.
For each PyStatsV1 exercise, keep this checklist nearby:
Question -> Data path -> Command -> Output -> Check -> Plain-language noteThat chain is the difference between a screenshot and a reproducible lesson.
Most early problems are environment problems, not statistics problems. Fix them one at a time and keep notes.
Check which command your system uses:
python --version
python3 --version
py -3 --versionUse one command consistently inside a project.
If installed packages disappear, activate the environment again:
source .venv/bin/activateOn Windows PowerShell:
.\.venv\Scripts\Activate.ps1Upgrade pip, then retry in a clean environment:
python -m pip install -U pip
python -m pip install -e .A failing test is useful evidence. Read the first failure, not the last screenful of output. Record:
Do not force the result to match your expectation. First check the input data, grouping variable, missing values, and assumptions. Then rerun the script from a clean baseline.
After the first proof works, continue in small steps.
Recommended next steps:
The project source is available at:
https://github.com/pystatsv1/PyStatsV1The LearnToProgram.ca MVP keeps the resource free while also proving a sandbox-only voluntary contribution path. Payment is not required for the download.
When you are unsure, return to the smallest reproducible proof. A smaller proof that you understand is more useful than a large analysis that you cannot explain.
PyStatsV1 software is MIT licensed. The full license text should be reviewed in the PyStatsV1 source repository.
This MVP PDF is educational material prepared for the LearnToProgram.ca portal. It summarizes a safe getting-started workflow and points learners toward the MIT-licensed PyStatsV1 software project.
A short MIT license reminder:
Permission is granted to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, subject to the conditions stated in the MIT License.This book does not remove the need to review the actual project license, source code, documentation, and tests.
Save evidence, not secrets. Do not paste API keys, payment credentials, private data, or student records into examples, screenshots, commits, or support messages.