4 th anz sra conference uncertainty analysis workshop keith r hayes csiro division of mathematical...

Download 4 th ANZ SRA Conference Uncertainty analysis workshop Keith R Hayes CSIRO Division of Mathematical and Information Sciences 28 th September 2009, Wellington

If you can't read please download the document

Upload: monica-mccarthy

Post on 27-Dec-2015

219 views

Category:

Documents


1 download

TRANSCRIPT

  • Slide 1
  • 4 th ANZ SRA Conference Uncertainty analysis workshop Keith R Hayes CSIRO Division of Mathematical and Information Sciences 28 th September 2009, Wellington
  • Slide 2
  • Overview Part I : Introduction and linguistic uncertainty uncertainty and its many sources identifying and treating linguistic uncertainty issues for qualitative risk assessment models and quantitative risk assessment Part II: Uncertainty analysis methods methods for representing variability methods for treating epistemic uncertainty pros and cons of different approaches Part III: Model structure uncertainty qualitative and quantitative approaches
  • Slide 3
  • Acknowledgements People who have helped: Simon Barry - statistics and general mentoring Scott Ferson - R functions for pba, helping me out of tight spots Mark Burgman case study material and elicitation Petra Kuhnert elicitation and pooling discussions Greg Hood R programming tips Funding that has helped: attendance and research partially funded by the Australian Centre of Excellence for Risk Assessment (ACERA)
  • Slide 4
  • Part I: What is uncertainty?
  • Slide 5
  • Why worry about uncertainty? Risk and uncertainty are intimately linked Risk occurs because the past and present can be uncertain, and the future is uncertain Reasons why you may want to address uncertainty perform an honest risk assessment (Burgman, 2005) ensure that the wheat remain separated from the chaff separate knowledge gaps from variability predict, measure, learn transparency Reasons why you may not want to address uncertainty takes more time and resources paralysis through analysis results may span decision criteria transparency
  • Slide 6
  • What is uncertainty? Some definitions: a degree of ignorance (Beven, 2009), a state of incomplete knowledge (Cullen and Frey, 1999) insufficient information (Murray, 2002) a departure from the unattainable state of complete determinism (Walker et al., 2003). Large number of taxonomies and classification schemes but basically: linguistic uncertainty, epistemic uncertainty variability
  • Slide 7
  • Terminology!
  • Slide 8
  • How do we represent uncertainty? Using language highly certain, low uncertainty Numerically probability imprecise probability Dempster-Shafer belief functions possibility measures ranking functions plausibility measures In practice probability far and away the most popular
  • Slide 9
  • Linguistic uncertainty
  • Slide 10
  • Ambiguity arises when words have more than one meaning and it is not clear which one is meant Context dependence caused by a failure to specify the context in which a term is to be understood: large scale escape Underspecificity occurs when there is unwanted generality: in a small percentage (generally