Teacher: Jukka Ranta (jukka.ranta at evira.fi) 

Scope: 5 op

Type: intermediate / advanced studies

Teaching: Lectures and computer class working.

Topics:This course consists of practical examples of Bayesian inference that can be implemented graphically by using so called 'DoodleBUGS' feature that is part of OpenBUGS (or WinBUGS) software. Each model is a DAG - a Directed Acyclic Graph which defines conditional probabilities between variables in the model. Once the graphical model is defined, the BUGS model code can also be printed out. However, we focus on the graphical models rather than BUGS-programming. The goal is to learn the basics with no requirement of extensive programming skills. Each example will introduce step-by-step some standard probability distributions when building the models. With a bottom-up-approach, we gradually get more familiar with the theory of Bayesian inference, and how it works with BUGS. The examples include practical problems with sampling problems (binomial-model), observed counts (Poisson-model), measurement data (normal-model), categorical data (multinomial-model) and regression models. We will work intensively hands on the practical examples with computers.

Prerequisites: Studies in probability, statistical inference, linear models, data-analysis with R-program. In statistic major degree requirements this course is scheduled for students who have already taken the courses 57045, 57046,  57703, 57705, 57701, 57714. A recommendation is that you have also taken a course which includes bayesian statistics, such as 57753, 57733, 57059, 57744. It is also possible to catch the relevant basics on bayesian statistics by self learning during this course (57745).