Lecture: Tue and Thu 10:05-11:20pm in Old Chem 025.
Lab (starting from Week 3): Mon 3:05-4:20pm in Old Chem 123.
Li Ma (Instructor), Email: li.maPENGUIN@dukePENGUIN.edu
Aihua Li (TA), Email: aihua.liPUFFIN@duke.edu
Don't forget to remove the arctic birds from the email addresses!
Mon: 4:30-5:30pm (Aihua) Old Chem 025.
Tue: 4-5pm (Aihua) on Zoom.
Wed: 4-5pm (Aihua) Old Chem 203B.
Fri: 2:30-3:30pm (Li) Old Chem 217.
All office hours share the same Zoom link, which will be distributed on Canvas.
Peter Hoff, A First Course in Bayesian Statistical Methods, 1st edition. (This text is free for download on SpringerLink when you are on Duke campus network or VPN.)
(optional) Christian P. Robert, The Bayesian Choice, 2nd edition. (This text is also free for download on SpringerLink when you are on Duke campus network or VPN.)
(optional) Andrew Gelman, John B. Carlin, Hal S. Stern, David B. Dunson, Aki Vehtari, and Donald B. Rubin, Bayesian Data Analysis, 3rd edition, ISBN-10: 1439840954. (This text is a good reference.)
(optional) James O Berger, Statistical Decision Theory and Bayesian Analysis, 2nd edition.
This course presents an introduction to the concepts and methods of Bayesian inference, with a focus on both modeling and computation.
Linear algebra, multivariate calculus based probability theory at the level of STA 230, mathematical statistics at the level of STA 432, and linear regression models at the level of STA 211.
Lecture notes will be distributed on Canvas.
| Week | Topics (might be adjusted as we go along) | Readings in PH |
| 1 | Inference using Bayes theorem, exponential family and conjugate models | Ch 1-3 |
| 2 | Decision theory, Monte Carlo approximation | Ch 4 |
| 3 | Rejection sampling and importance sampling | |
| 4 | Non-informative priors, univariate normal model | Ch 5 |
| 5 | Gibbs Sampling, Metropolis-Hastings, MCMC diagnostics | Ch 6, 10 |
| 6 | Multivariate normal model, Midterm exam | Ch 7 |
| 7 | Missing data and imputation, hierarchical models | Ch 7-8 |
| 8 | Bayesian linear models, hypothesis testing, model choice | Ch 9 |
| 9 | Latent variable models | |
| 10 | Final, advanced Bayesian computation | |
| 11 | Bayesian nonparametrics | |
| 12 | Generative models, student presentations | |
| 13 | Student presentations |
Weekly homeworks will be graded on a 4-point scale (Excellent, Good, Fair, and Poor). Both an Excellent and a Good will give you full credit for grading purposes. You must show your work to receive credit. Late homeworks will be accepted, but will incur a one-level grade penalty for each 24-hour period it is late (starting from the minute past the deadline). The lowest homework grade will be dropped. Homeworks are to be released and submitted on Gradescope.
There will be a total of 9 to 10 weekly lab sessions on Mondays starting from the third week. The main purpose of the lab is to provide you with a training in implementing Bayesian methods in R. Occasionally the lab might be used for reviews. Grading on the labs is only based on attendance and participation.
There will be two exams (one midterm and one final), both in-class and close-book/notes. For the exams, you may bring a calculator and a letter sized cheat-sheet (1-sided for the midterm and 2-sided for the final).
If you have an unchangeable conflict with either of the exams, you must submit the appropriate university online form (NOVAP, RHoliday) and arrange with me at least one week prior to the scheduled exam. The make-up exam in such cases will typically occur before the scheduled one. If you miss the midterm due to short-term illness, you should submit the online Incapacitation Form before the exam, and email me before the exam. In such cases your final exam will count for both exams.
There will be a course project due on the last day of class (November 26) and it should be completed individually. Each student will give a 10-min presentation on their project.
The course grade will consist of weekly homework (10%), weekly labs (10%), the midterm (30%), the final (35%), and the course project (15%). All grades are reported on Gradescope.
Small (online or in-person) study groups are encouraged.
Discussions on the homework problems are allowed, but you must write down your own solutions independently.
Each student is committed to Duke's Community Standard. No form of academic dishonesty will be tolerated. Some examples include cheating, plagiarism, and lying about illness or other reasons for absence. Violations of the standard will result in an automatic F of this course and will be reported to the Office of Student Conduct.