Bayesian networks with forensic and other applications2010HT
|
|
Course plan
No of lectures
Preliminary 16 (including supervised exercises)
Recommended for
PhD students within areas of forensic, natural and medical science, who need training in probabilistic reasoning and decision theory.
The course was last given
Goals
The course provides understanding of basic concepts and teaches skills employed
in the statistical evaluation of physical evidence. Forensic evidence will be
used as examples. However, the students are strongly encouraged to bring
projects/problems applicable to Bayesian network analysis also from other
scientific areas such as medical genetics, ecology, evolutionary biology and
clinical medicine.
Upon completion of the course, the participants should be able to:
- use knowledge about basic probability models and graphical network analysis
for evaluation of physical evidence (eg forensic and medical genetics,
(forensic) chemistry, classical criminalistics (shoe prints, glass
fragments..), other areas such as ecology etc.
- display good understanding of major principles for formulation of hypothesis
and likelihood ratios and the construction of Bayesian networks
- estimate likelihood ratios from empirical data and interpret their values on
ordinal scales of conclusion
- use standard software for Bayesian networks
Prerequisites
Basic knowledge about probability theory corresponding with an introductory course on statistics or mathematical statistics.
Organization
The course is given as a semi-distance course according to the following template: 4 weeks of concentrated lectures, seminars and computer exercises and practicals (3-4 days per week), interspaced with 3 weeks of individual studies.
Contents
The course content comprises:
- the concept of probability and Bayes’ theorem on odds form
- Bayesian hypothesis testing, hierarchy of propositions and likelihood ratios
- graphical models for probabilistic reasoning and Bayesian networks
- different types of evidence: DNA, illicit drugs and trace evidence
- transfer evidence, combination of evidence and pre-assessment
- sources of errors, fallacies and forensic interpretation on ordinal scales
- training in a computer package for Bayesian networks
- practical examples/applications from the course delegates
Literature
Taroni, F., Aitken, C, Garbolino, P. & Biedermann, A. Bayesian Networks and Probabilistic Inference in Forensic Science. Chichester: Wiley, 2006. ISBN 0470091738
Lecturers
Anders Nordgaard, Linköping university and Swedish National Laboratory of
Forensic Sciences
Petter Mostad, Chalmers university
Examiner
Anders Nordgaard
Examination
To pass, each student should show that he/she is able to apply the contents of the course in an individual project with a subject chosen from his/her own area of expertise. The project work should be presented as a written report and orally at a final seminar.
Credit
6
Comments
The maximum number of participants is 30.
The course is given in cooperation with the Center for Forensic Science (the
National Laboratory of Forensic Science - SKL , the National Board of Forensic
Medicine - RMV) and Linköping University (LiTH and HU).
Page responsible: Director of Graduate Studies