University of Technology, Sydney

Staff directory | Campus maps | Newsroom | What's on

UTS-CSIRO Workshop on Change-Point Analysis: Models, Methods and Applications


 "Change-Point Analysis: Models, Methods and Applications"


Principal  speaker:


Albert  Shiryaev (visiting Professor at UTS,  supported by the QFRC)


Speakers from CSIRO: Dr.  Ross Sparks and Dr.  Adrien  Ickowicz


 


Time: 2:00pm 5:30pm, Thursday, February 14, 2013


Venue: Room 11, Level 16, UTS Tower Building, 15 Broadway, Sydney


Please RSVP before February 12, 5pm to the coordinator of the workshop


Dr. Alex Radchik (Alex.Radchik@uts.edu.au) for catering purposes.


 


Albert   Shiryaev  is known for his work in probability  theory, statistics and financial mathematics. He has been working in Steklov Mathematical Institute since 1957. He earned his candidate (PhD) degree in 1961 (Andrey Kolmogorov  was his advisor) and a doctoral degree in 1967 for his work "On statistical sequential analysis". He is a Head of the Department of Probability of Moscow State University. Albert was elected a full member of the Russian Academy of Sciences in 2011. Albert Shiryaev’s main contributions are in problems of quickest detection of change-points, nonlinear filtration, stochastic differential equations, stochastic optimization. He is an author of numerous textbooks and monographs including "Optimal Stopping Rules"and "Essentials of Stochastic Finance".


 


 


Ross Sparks is a Leader of Health Informatics Group in Division of Mathematics, Informatics and Statistics (CSIRO). His key skills are in health surveillance and monitoring of public health outcomes,  applied multivariate  analysis, spatial temporal modelling of health data, epidemiological methods.


 



Dr. Adrien Ickowicz is Postdoctoral Fellow - CSIRO Mathematics, Informatics and Statistics




-----------------------------------------------------------------------------------------------------------------------------------------------------------------------


Abstracts of presentations:



 


Albert  Shiryaev, 2-3.40pm


 


Optimal  stopping problems for a Brownian motion with disorder on a finite interval,  with application to trading of bubbles


 


 


We present a series of new results on the Bayesian optimal stopping of both theoretical and applied interest. The applications which we consider concern the problem of the trading of bubbles on financial markets.


Our model for bubbles is described by the process X = (Xt)tT with


 


Xt = µ1t + (µ2 µ1)(t θ)+ + σBt


 


and by the geometrical  process S = (St)t with



 


 


 




  with some functions f (i).



Using some  empirical data we demonstrate  that  the model suggested provides good (profitable) exit trading points.


The presentation is based on two recent papers:


[1] A.N.Shiryaev, M.V.Zhitlukhin, Optimal stopping problems for a Brownian motion with a disorder on a finite interval (2012).


[2] A.N.Shiryaev, M.V.Zhitlukhin,  W.T.Ziemba, When to sell Apple and the Nasdaq? Trading bubbles with a stopping rule model (2012).


 


 


 


 


Ross Sparks, 4-4.45pm


 


"Some Approaches to Finding Spatio-Temporal Outbreaks of


Diseases"


 


The SATSCAN software offers a number of capabilities for detecting  disease clusters  (e.g., see Kulldorff and Nagarwalla,  1995, Kulldorff, 1997, Kulldorff,


2001, Kulldorff et al., 2005). Takahashi et al (2008) extended the scan approach to detect single clustered outbreaks of various shapes, but their approach is based on (Kulldorff, 2001)’s scan statistic. Sparks (2010) demonstrated that (Kulldorff, 2001)s scan statistic is inappropriate for outbreak detection when the population size is unknown - the situation considered in this talk.


The disease outbreak geography can vary in shape, size and number of disconnected outbreak regions. Robust  methodology is required that  will adapt to this variation in outbreak characteristics. The traditional approach of aggregating  disease counts over all of a target region and monitoring the total volume fails to exploit the fact that diseases cluster spatially, and as such the aggregation dampens the influence of the outbreak delaying finding it. A methodology that exploits the spatial clustering is required.


In this talk, we compare incidences to expected values based on historical trends, where the expected values can be space-time dependent and whose in-control behaviour is assumed  to change  slowly with  time. The aim of the talk is to deliver plans that are robust at detecting multiple clustered outbreaks of varying shapes and sizes quickly. Such disease outbreaks are commonly caused by several people being infected  overseas or during transit while returning to their home country. These infected


Date:
14 February 2013
Time:
14:00 - 17:30
Location:
City - Broadway CB01 Tower
Audience:
All Welcome
Contact:
Alex Radchik

Back to list of archived events