Aggregation, Inference and Rare Events in the Natural and Socio-Economic Sciences
Two Day Research Workshop: Aggregation, Inference and Rare Events in the Natural and Socio-economic Sciences
Warwick Mathematics Institute
Warwick Centre for Complexity Science
17-18 May 2012
Organisers: Colm Connaughton (Warwick Mathematics Institute and Warwick Centre for Complexity Science) and Nick Watkins (British Antarctic Survey)
The aggregation of random fluctuations in complex systems is a problem with aspects as abstract as the renormalisation group and as concrete as the risk industry. Classical statistics has given us the central limit theorem, describing the flow, under aggregation, of light-tailed fluctuations towards the Gaussian limit. In this context extreme events are rare, and are handled in the correspondingly mature framework of extreme value theory .
However, laboratory critical phenomena, fluid turbulence, and a wide range of socio-economic systems are increasingly recognised as giving rise to heavier-tailed distributions of fluctuations, in which "extreme" events are correspondingly much more common. Much progress has been made, notably through the use of additive models with alpha-stable (“Levy”) distributions, or by multiplicative cascade processes, but many important open problems remain.
This meeting focuses on three timely aspects of this broad topic:
- How can we generalise the idea of limit theorems to the aggregation of dependent, heavy-tailed fluctuations ? We consider for example experimental systems of finite size such as confined turbulence, finite sized models, or the aggregated stock prices in a given market sector.
- What is best practise for inference in complex systems ? We are interested in the problem of developing methods for distinguishing candidate models which are statistically robust in the presence of imperfect data but still rich enough to encompass the observed distributions.
- How does the type of aggregation seen in a complex system condition its extreme behaviour ? Here we are interested both in the extreme behaviour of additive and multiplicative models and also in the links between extremes and sums in highly correlated, hierarchical systems.
- Steve Bramwell (UCL)
- Cosma Shalizi (Carnegie-Mellon)
- Maxime Clusel (Montpellier)
- Iddo Eliazar (Holon Institute of Technology)
- Marcin Magdziarz (Wroclaw)
- Ola Løvsletten (Tromso)
- Rainer Klages (QMUL)
- Lenny Smith (LSE)
- Simon N Wood (Bath)
- Tim Graves (Cambridge)
- Christos Mitas (Risk Management Solutions)
- Gunnar Pruessner (Imperial)
- Bruno Goncalves (Northeastern)
Programme is available here [contains pdf versions of most of the presentations].
Deadlines and registration:
You can register for this event here. (Please scroll down until you find the appropriate event and click the "continue" button.)
Due to restrictions on the room capacity, registration will close if we reach capacity.