3 credits
Fall 2025 Lecture Distance Learning Upper Division
The course focuses on two fundamental aspects in computational statistics: (1) what to compute and (2) how to compute. The first is covered with a brief review of advanced topics in statistical inference, including Fisher's fiducial inference, Bayesian and frequentist methods, and the Dempster-Shafer (DS) Theory. The second is discussed in detail by examining exact, approximation, and interactive simulation methods for statistical inference with a variety of commonly used statistical models. The emphasis is on the EM-type and quasi-Newton algorithms, numerical differentiation and integration, and Markov chain Monte Carlo methods.
Learning Outcomes1Master commonly used numerical matrix operations, including the sweep operator, Cholesky decomposition, eigenvalue decomposition, and single value decomposition.
2Understand statistical thinking in development of interactive numerical methods.
3Apply optimization algorithms such as quasi-Newton and conjugate gradient methods.
4Implement EM-type algorithms for maximum likelihood estimation when expanded complete-data models are available, including commonly used statistical models.
5Create random number generators.
6Implement the Gibbs sampler and Metropolis-Hastings algorithms for Bayesian data analysis.
Prerequisites
Restrictions
GPA by professor
made by Purdue students.