Bayesian Analysis with Python, 2nd Edition

Bayesian Analysis with Python, 2nd Edition is an introductory book on Bayesian statistics and probabilistic programming from a hands-on programming perspective; the original title is Bayesian Analysis with Python, with the subtitle "Statistical Modeling and Probabilistic Programming with PyMC3 and ArviZ." The book does not rely on elaborate mathematical derivations; instead, using the probabilistic programming library PyMC3 and the Bayesian model exploration and analysis library ArviZ as tools, it guides readers step by step from Bayesian inference to code. The content proceeds in three stages—foundations, modeling, and extensions: Chapter 1 starts from the relationship between statistics and modeling, explaining the meaning of probability, single-parameter inference, how to report posterior results, and posterior predictive checks; Chapter 2 moves into probabilistic programming proper, introducing the PyMC3 modeling workflow, posterior summaries, Gaussian distributions and robust inference, comparisons between groups, and hierarchical models and shrinkage; Chapters 3–4 cover linear regression, robust regression, hierarchical regression, polynomial regression, multiple regression, and generalized linear models such as logistic regression, softmax regression, Poisson regression, and zero-inflated Poisson models; Chapters 5–7 discuss model comparison (information criteria, WAIC, Bayes factors, regularizing priors), finite and continuous mixture models, and Gaussian process regression, classification, and Cox processes; Chapter 8 explains inference engines, including non-Markovian and Markovian methods and sample diagnostics; Chapter 9 points toward further study. Compared with the first edition, the second edition has been extensively rewritten and reorganized; most figures are generated directly from the book's code, and full-color printing is used to better present code and visualizations. Each chapter ends with exercises, and the accompanying code can be downloaded from the Asynchronous Community (epubit.com) site, making the book suitable for systematic study by readers who want to use probabilistic thinking to solve real data analysis problems.