Syllabus
1. Basics of Bayesian statistics, likelihood, duality of model and data.
2. Principle of Markov Chain Monte Carlo (MCMC) and basics of OpenBUGS and JAGS.
3. Simple Bayesian models, bestiary of probability distributions and their implementation in BUGS language.
4. Generalized linear models - linear regression, logistic regression, Poisson regression, ANOVA etc., all that in BUGS.
5. Hierarchical (mixed-effect, multilevel) models, random effects vs. fixed effects, latent variables, complex models, informative vs. non-informative priors.
6. Time series analysis, autocorrelation function, density dependence, random walks.
7. Modelling spatial and geographical data, spatial autocorrelation, GeoBUGS module.
8. Model selection, model evaluation, information theory criteria, handling uncertainty, Bayesian credible intervals, prediction intervals.
Annotation
The course is an introduction to modern applied Bayesian statistics. Its aim is to show that one can do statistics outside of the classical frequentist categories, and that statistics can work as a modular kit in which several simple blocks can be used to analyze problems of any complexity.
The course will emphasize the practical adavantages of Bayesian statistics rather than the theoretical or philosophical ones. We will use mainly ecological examples but the methods are universally applicable throughout the whole biology.
Particularly, participants will learn to specify, fit and evaluate models in BUGS language (OpenBUGS, JAGS). The course assumes basic knowledge of R (i.e. "I can launch R, load data into R, I can do a simple regression model" and so on).
Basic programming skills will be advantageous (but not critical). The course will have a form of a 3-4 days of intensive seminar-workshop.
The course can be taught either in Czech or English, depending on the language skills of the participants.