We are used to classical statistics, yet in many important contemporary applications, such as the estimation of parameters in machine learning, a second branch of statistics based on Bayesian methods is very attractive. Therefore, a book that swiftly and elegantly introduces this area will interest readers. A contemporary textbook should include not only a clear description of the field and related theory, but also an introduction with software examples. This is the case with Hooten and Hefley’s book. Here, the authors deal with R, a well-known language, to help readers process statistical reasonings.

The material is divided into very small pieces of information, thus the chapter granularity is very large. I do like this approach, since it enables readers to comprehend a self-contained fragment within a single reading session. The book is organized into five main sections:

• Section 1, “Background,” covers the most important ideas related to Bayesian modeling: relationships and differences with classical probability/statistics theory; numerical integration; Monte Carlo methods, including Markov chain Monte Carlo; and importance sampling.
• Section 2, “Basic Models and Concepts,” illustrates the mentioned fundamental data with parameter estimation for various basic distributions (Bernoulli-Beta, normal-normal, normal-inverse gamma, normal-normal-inverse gamma).
• Section 3, “Intermediate Models and Concepts,” deals with practical inference scenarios (mixture models, linear regression, posterior prediction, model comparison, regularization, Bayesian model averaging, time-series modeling, and spatial models).
• Section 4, “Advanced Models and Concepts,” covers inference on less typical values (quantile regression, hierarchical models, binary regression, count data regression, zero-inflated models, occupancy models, and abundance models).
• Section 5, “Expert Models and Concepts,” looks at various complex approaches to show how Bayesian statistics can benefit from data gathered in a more advanced way (integrated population models, spatial occupancy models, spatial capture-recapture models, spatio-temporal models) and finishes with the Hamiltonian Monte Carlo method.

As can be seen from the above overview of topics, this almost 600-page book is really rich in material. An attractive aspect of the discussion is that it follows a bottom-up approach, that is, it bases more advanced aspects on the ones presented before from scratch. This way, Houten and Hefley present a work that readers can use to get started with Bayesian methods and obtain quite a high level of excellence in the field. Yet it is necessary to note that readers with previous statistical training will more easily understand how the ideas are introduced. While the mentioned software examples in R are very helpful and allow the reader to more easily understand the presented formalism, computer scientists and networking readers may be unhappy about the lack of illustrations related to their fields. This stems from the authors’ backgrounds, which are not in computer science. However, I do not perceive this as a drawback, since the book gives a good overview of statistics and it is the reader who is responsible for applying it in their field of interest.

More reviews about this item: Amazon