贝叶斯数据分析

Bayesian inference is the process of fitting a probability model to a set of data and sum- marizing the result by a probability distribution on the parameters of the model and on unobserved quantities such as predictions for new observations. In Chapters 1–3, we in- troduce several useful families of models and illustrate their application in the analysis of relatively simple data structures. Some mathematics arises in the analytical manipulation of the probability distributions, notably in transformation and integration in mult iparameter problems. We differ somewhat from other introductions to Bayesian inference by emphasiz- ing stochastic simulation, and the combination of mathematical analysis and simulation, as general methods for summarizing distributions. Chapter 4 outlines the fundamental con- nections between Bayesian and other approaches to statistical inference. The early chapters focus on simple examples to develop the basic ideas of Bayesian inference; examples in which the Bayesian approach makes a practical difference relative to more traditional approaches begin to appear in Chapter 3. The major practical advantages of the Bayesian approach appear in Chapter 5, where we introduce hierarchical models, which allow the parameters of a prior, or population, distribution themselves to be estimated from data null iparameter problems. We differ somewhat from other introductions to Bayesian inference by emphasiz- ing stochastic simulation, and the combination of mathematical analysis and simulation, as general methods for summarizing distributions. Chapter 4 outlines the fundamental con- nections between Bayesian and other approaches to statistical inference. The early chapters focus on simple examples to develop the basic ideas of Bayesian inference; examples in which the Bayesian approach makes a practical difference relative to more traditional approaches begin to appear in Chapter 3. The major practical advantages of the Bayesian approach appear in Chapter 5, where we introduce hierarchical models, which allow the parameters of a prior, or population, distribution themselves to be estimated from data null