Quality and Innovation

While Python is my go-to language for things like NLP, I usually use R for everything else. After spending a solid long weekend with Martin’s new book “Bayesian Analysis with Python” I can confirm that this book will be just what ONE audience needs, but may be less useful for others. Are you in the target audience?? As a gentle introduction to Bayesian approaches for people who are well versed in intro statistics and have a solid foundation in Python, it’s perfect. But if you’re missing that mathematical statistics background (or if you’re rusty on Python and don’t have a development environment already set up & ready to run code) this book may present a struggle.

🌟 As a result, I give this five stars for the target audience and four for the other audiences.

The writing is clear and easy to follow, but sometimes encourages you to “review the code for understanding” where the author had a chance to explain each of the lines of code (but didn’t). As a result, you’ll need to allocate more time to practice and put in the diligence to make sure concepts “click”. The book also assumes that the reader has a fundamental understanding of distributions and mathematical notation, which may not be the case for all programmers or data analysts. As a professor this would have been a great book to use from an introductory Bayesian methods course for juniors or seniors in STEM with at least one or two semesters of Python. For this group, the book is particularly strong, because it takes a computation-first approach but fills in the gaps with just enough theory.

Highlights include:

– Simple discussion on ROPE and loss functions that is valuable
– Good discussion about how to do linear regression the Bayesian way (hint: all parameters treated as priors)
– Some interesting mixture models using the Palmer Penguins dataset
– Best part was the MCMC with Metropolis-Hastings to calculate the value of pi, which I could easily pair with less sophisticated MC methods to show the difference

☑ DO buy this book if you have a solid foundation in Python (and a Python environment already set up) and want to spend a few weeks (or a couple months) expanding your understanding into building and running simple Bayesian models. If you have the time to spend, this will deepen your understanding.

⚠ DO NOT buy this book if you are a programmer who needs to start building Bayesian models at work for a deadline next week 🙂 You’ll need a little more time and depth before you can knock out a project at work. But if you’ve got that time, this is a great resource for beginning your understanding of Bayesian methods using Python!

#bookreview #python #bayesian