Part XVII — Statistical Computing

The numerical workhorses behind modern inference: optimization and integration, the EM algorithm, and the MCMC samplers that make Bayesian computation practical.

Topics

Topic	Focus
Numerical Optimization	Gradient-based and derivative-free optimization for fitting models
Numerical Integration	Quadrature and the integrals simulation replaces
Expectation-Maximization Algorithm	Iterative maximum likelihood with latent variables
Markov Chain Monte Carlo	Sampling from intractable posteriors by running a chain
Gibbs Sampling	MCMC by cycling through conditional distributions
Metropolis–Hastings	The accept–reject rule behind general MCMC