Next week (Tuesday 16 Oct) 5.30 for 6-7pm.
Department of Management Science, University of Strathclyde, 40 George
Street, Glasgow.
Modelling subjective probabilities with vines by Prof Tim Bedford
This talk will be a general introduction to the reasons for using
subjective probabilities, problems and limitations of the concept, and new
methods for eliciting dependencies. First of all I will talk a bit about
the meaning of subjective probability, in order to clarify what we can do
and more importantly what we cannot do with subjective probability. With
some simple examples (motivated by technology risk assessment) I will show
that different subjective models are plausible and that these can give
quite different answers. These examples show how important it is to be
clear about the objectives of the modelling. An important application of
subjective probability models in the recent past has been in modelling the
effect of state-of-knowledge uncertainties in consequence models. Often it
turns out that dependencies between uncertain factors can be important in
driving the output of the models. Various methods have been developed to
model these dependencies, in particular simple correlation methods which
have been built into simulation packages such as @Risk. There are however
serious problems with these models, which have motivated the development of
vine dependent models. While vines are slightly more complex
mathematically, they are more suitable for applying expert elicitation.
|