PNS0157 Understanding and Implementing Bayesian Modelling: a Course from Beginnings to Hierarchical Complexity, 5.0 Credits
No Level Indicated
Pass / Failed
The requirements for attaining different grades are described in the course assessment criteria which are contained in a supplement to the course syllabus. Current information on assessment criteria shall be made available at the start of the course.
Admitted to PhD studies. The course will assume no prior knowledge of Bayesian statistics but will require that students are able to use R (e.g. creating sequences, lists, importing data) and have a basic understanding of R programming principles (e.g. loops, indexing etc).
Students will have a grounded understanding of:
(1) how Bayes rule is derived from the principles of probability,
(2) the advantages of Bayesian analyses and when they should be used,
(3) the probability distributions needed for linking their data to ecological models (Gaussian, Poisson, Binomial, Gamma, Beta, Bernoulli, Multinomial),
(4) how statistical models are constructed and described,
(5) how to use hierarchical modelling to account for structures in the data and to ask specific ecological questions that are unavailable in other modelling approaches,
(6) how to interpret and communicate information contained in the model results, and
(7) how to use R and JAGS to do all of these things
The students are encouraged to work both independently and within small groups to help problem-solve modelling issues. This encourages the student to understand how to solve problems when working under different working conditions and encourages communication and collaboration. Both teachers circulate around the room offering encouragement, clarifying concepts and helping students solve problems as they arise.
Lectures are given around the key concepts that form the basis of the exercises. These key concepts given in lectures are:
1. Bayes rule
4. Prior knowledge
6. General Linear Models
7. Describing Models
8. Using JAGS
9. Making Predictions
10. Interactions, Collinearity and Scaling
11. Non-linear Models
12. Hierarchical Models
13. Adding Prior Information to Models
15. Posterior Predictive Checks (is my model ok?)
16. Adding Observational Uncertainty
17. State-Space (time-dependant) Models
Formats and requirements for examination
The course will be structured around the principles of formative assessment. This means that there will be no end-of-course assignment or exam to complete: rather the teachers will continuously assess the students during the course. The advantage of this approach is that students who understand material quickly can be given more challenging exercises to progress with, while students who are having trouble understanding key concepts can be given more time and additional exercises to help them understand before moving on. This ensures all students work to their ability to gain the most out of the course. Thus passing the course will be determined by attendance and work ethic, rather than an arbitrary achievement level.
The course organiser and primary teacher will be Matt Low and the teaching assistant will be Malin Aronsson, both Dept. of Ecology.
Department of Ecology