Bayesian graphical modelling
It is a powerful methodology for visualizing and analyzing complex epidemiological data,it is ideally suited for risk factor studies and analysis of questionnaire data. This technique may reveal far more about complex disease systems than standard approaches when dealing with many inter-related variables, as is common in farm survey data.
Workshop
It is largely computer based and uses R. Some familiarity with R would be advantageous but not essential, as would previous experience of basic epidemiology and statistical modelling, such as linear regression techniques.
Instructor
Dr Fraser Lewis is an Applied Statistician in the Section of Epidemiology, Vetsuisse. Fraser has over 10 years experience working with epidemiological data following degrees in Statistics and a PhD in Mathematics. He has recently received funding from the Swiss National Science Foundation to develop Bayesian network models for applications in epidemiological research, in collaboration with the Farm Surveillance Unit within the Public Health Agency of Canada.
Date, Venue and Registration
The workshop will be held on 15-17 February at the Camden Campus of The University of Sydney (425 Werombi Road, Camden NSW 2570).
This course is being run on a cost-recovery basis. The course fee is $150 for students and $300 for non-students. This fee includes the course materials, morning and afternoon tea/coffee and lunch for each day.
To register for the workshop please contact Dr. Navneet Dhand: navneet.dhand@sydney.edu.au. PROGRAM Day 1 AM Introduction to Bayesian Graphical Models - Published examples of BGMs - Differences between BGM and regression modelling - Hands-on exercises in R PM Fitting Bayesian Graphical Models to Data - Some basic probability theory - Directed acyclic graphs - Bayesian Estimation and Parameter Learning - Hands-on exercises in R Day 2 AM Structure Learning - Challenges of model selection - Goodness of fit metrics for BGMs - Using R to find the best model for a given dataset - Hands-on exercises in R PM Parametric Bootstrapping in Model Selection - a simple regression example - using JAGS/WinBUGS to generate bootstrap samples - Hands-on exercises in R and JAGS Day 3 AM Order Based Structure Learning - Exact algorithms - Finding maximal posterior structures - Hands-on exercises in R PM Other types of BGMs - Gaussian - Mixed conditionally Gaussian - Additive networks - Hands-on exercises in R Download Information Flyer(pdf) ---------------------------- Important Notes: 1 The pace and content of the workshop is subject to change and may be altered depending on the experience of the participants. 2 Participants are welcome to bring along their own data for use in some of the exercises. | 
Expand all