Applied Bayesian data analysis, 7.5 credits, FOR046F
Education level: Doctoral level
Course name: Applied Bayesian Data Analysis
ECTS/HP: 7.5
Course code: FOR046F
Educational level: third-cycle course.
Eligibility requirements: The course is open for students admitted to third-cycle studies at LTU.
Entry requirements: Statistical knowledge corresponding to the basic course S0001M or S0008M or equivalent course. Programming experience in e.g., R, MATLAB, Python or similar. The primary tool for this course will be R.
Course content:
The course explores various methods in Bayesian data analysis for drawing conclusions from empirical data, incorporating previous experience and knowledge (if available). In contrast to classical (frequentist) data analysis, Bayesian approaches offer more robust and comprehensive inference. Specifically, the course will cover:
- Bayes’ Theorem. Differences between classical (frequentist) and Bayesian inference.
- Posterior inference: summarizing posterior distributions, credible intervals, posterior probabilities, posterior predictive distributions, and data visualisation.
- Gamma-Poisson, beta-binomial and normal conjugate models for data analysis.
- Bayesian regression analysis and analysis of variance (ANOVA).
- Use of simulations for posterior inference. Simple applications of Markov chain-Monte Carlo (MCMC) methods and their implementation in R and Stan.
- Bayesian cluster analysis.
Learning outcomes: The primary goal is to equip participants with the tools to apply and comprehend Bayesian models in real-world scenarios. Upon successful completion, participants will possess the knowledge and skills to:
- Explain in detail the Bayesian framework for data analysis, emphasizing its benefits and flexibility compared to the frequentist approach.
- Construct flexible Bayesian models using likelihood and prior functions.
- Demonstrate the role of the prior distribution in Bayesian inference, expressing the use of non-informative priors and konjugerade priors.
- Develop, elucidate analytically, and implement both single and multiparameter probability models within the Bayesian framework.
- Fit hierarchical models and provide comprehensive technical specifications for these models.
- Grasp the concept of Bayesian linear regression.
- Perform Bayesian computation with R and Stan.
Course methods: Teaching consists of lectures and seminars. Mandatory assignments are made continuously during the course. The course requires active participation during the seminars where individual assignments are presented.
Examination form: Written individual assignments, active participation in problem seminars.
Grading scale: Pass/Fail.
Course literature:
- John K. Kruschke (2014). Doing Bayesian Data Analysis: A tutorial with R, JAGS, and Stan. Elsevier Science, 2nd edition.
- Andrew Gelman, John Carlin, Hal Stern, David Dunson, Aki Vehtari, and Donald Rubin (2022). Bayesian Data Analysis, CRC Press, 3rd edition.
Education cycle: Sp 3-4, 2024.
Course is given periodically: Yes.
Send application to: Mohammad Ghorbani, Email: mohammad.ghorbani@ltu.se
Deadline for application: January 5.
Course open for application by doctoral students admitted to other universities than LTU:
No.
Limited number of students: The first 20 applicants.
Tuition:
Contact person: Mohammad Ghorbani, Email: mohammad.ghorbani@ltu.se, 0920-492197
Examiner: Peter Wall
Updated: