UNIVERSITY OF WATERLOO Department of Statistics and Actuarial Science SPEAKER: Dr. Geoff Chaplin TOPIC: "Correlation and Basket/CDO Pricing" OUTLINE 1. Background: Products, Concepts, Rating and Spread, Default, Mathematical Correlation, Credit Correlation, Calibration, The Risk-Free Rate and Bond Pricing 2. Single name CDS: simplified and full model 3. Simplified portfolio models: BET, Binomial, multi-Poisson, and Copula 4. Full portfolio model 5. Comparison of Results 6. Conclusions Geoff Chaplin is Head of Structured Credit Derivatives at ABN AMRO Bank NV, based in London. Previously, he worked for Nomura International in a similar role, and prior to that was with HSBC Markets (government bonds) and his own consultancy company. He has worked in London since 1980 in fixed income, trading and research. Geoff was an active member of the Fixed Income Working Party of the Institute and Faculty of Actuaries, and was a visiting professor at the University of Waterloo from 1988 to 1996. He has published articles in Risk magazine, the British Actuarial Journal, several professional magazines and spoken at numerous conferences. He graduated from Peterhouse, Cambridge, as a titled scholar of mathematics, did his doctorate at Oxford in quantum gravity, and then qualified as a Fellow of the Faculty of Actuaries. He can be reached at 44 20 7678 3965 or geoff.chaplin@uk.abnamro.com DATE: Wednesday May 30, 2001 TIME: 10:30 am PLACE: MC5158 Coffee and cookies will be available after the talk in MC6123 Nandanee Basdeo Statistics and Actuarial Science University of Waterloo Waterloo Ontario Canada N2L 3G1 Phone - 888-4567 X5025
McMASTER UNIVERSITY CLINICAL EPIDEMIOLOGY & BIOSTATISTICS ROUNDS Here are a few highlights of the upcoming Rounds Presentation for this week: May 31st, 2001 - 1:45 to 2:45 pm - HSC-4E20 Dr. Jean-Paul Collet, MD, PhD, Associate Professor, McGill University Associate Director of Medical Research, Lady Davis Institute Director, Randomized Clinical Trials Unit, Jewish General Hospital Title: "Web integrated support system for clinical research" Abstract: Background: Accurate and timely documentation of data is vital to the successful completion of clinical and epidemiologic research. Similarly, a clear and comprehensive approach to data management is a salient aspect of quality control in research. We developed a web-site data entry and management program to optimize research quality. System: Data collection documents are prepared for web-site compatibility. Oracle Database is then set up to check errors. The system is also programmed for double data entry. Error sheets for corrections are edited and designed for electronic transmission. The activity of each site (recruitment, missing visits, forms or data) is automatically updated and transmitted to the site; a study summary is also available for all investigators. The system may be used for planning follow-up visits and rapid detection of problems (alert). It is fully compatible with SAS, Excel and other software. The whole system is operated with strict safety procedures: encrypted data, back-up procedures, electronic lock, etc. Results: The web-site data management program has been used for conducting several multicenter RCTs and pharmacoepidemiology studies. Investigators found that this system facilitates the following research tasks: i) immediate monitoring of data quality, ii) verification of adherence to protocol, iii) monitoring of compliance with randomization, and iv) promoting timely communication with study site coordinators. Net benefits are: improving data quality, quick access to a clean database and facilitation of field monitoring. Conclusions: Web-site data management provides investigators with an efficient, innovative and advantageous alternative to traditional data management methods used in research. "This event is an accredited group learning activity as defined by the Maintenance of Certification Program of The Royal College of Physicians and Surgeons Canada"
UNIVERSITY OF WATERLOO Department of Statistics and Actuarial Science SPEAKER: Dr. Samuel Wong University of Science and Technology, Hong Kong TOPIC: "Generalized Continuous Threshold Autoregression and Hinging Hyperplanes " ABSTRACT The continuous threshold autoregression (CTAR) is a piecewise linear autoregressive model which is continuous everywhere. While CTAR restricts the threshold boundaries to be perpendicular to the axis of the delayed variable Xt-d, the generalized continuous threshold autoregression (GCTAR) provides flexibility to the threshold boundaries such that they can be oblique to the natural axes of the lagged variables. A close relative of GCTAR in machine learning area is hinging hyperplane which is equipped with fast estimation algorithm. In this article, a trust-region variant of the algorithm is found to be efficient for the conditional least squares estimation of GCTAR. The asymptotic normality of the estimates are obtained via empirical process theory. Simulated data are used for illustration. Also, GCTAR is applied to the quarterly unemployment rates and US Real GNP. The results obtained are quite different from those of CTAR. DATE: Thursday May 31, 2001 TIME: 3:30 pm PLACE: MC5158 Coffee and cookies will be available after the talk in MC6123 Nandanee Basdeo Statistics and Actuarial Science University of Waterloo Waterloo Ontario Canada N2L 3G1 Phone - 888-4567 X5025
Speaker: |
Dr Alison Burnham |
Topic: | |
Date: |
Wednesday June 6, 2001 |
Time: |
2:00 - 4:30 p.m. |
Location: |
Royal Bank Training Centre Take the King Subway exit on the north-west side of King and Yonge. For those who drive, please don't park on any main streets (King, Yonge, Bay or Adelaide); cars will be towed away from 3:00 to 6:00 p.m. |
SPEAKER: |
|
TITLE: |
"Variable selection and a simple alternative to PLS in chemical calibrations" |
DAY: |
Monday, June 18, 2001 |
TIME: |
11:00 a.m. |
PLACE: |
BSB-108 |
We present some recent results in wavelength selection and calibration. The selection methods that we use work in real time and are 100's of times faster than the popular genetic algorithms. Our methods produce comparable results. We also show that, contrary to claims in the chemical literature, there is no clear choice, typically, between direct or inverse calibration when some calibration ingredients are unknown. Finally, we present our parallel calibration method. It is the simplest modern calibration method, from all points of view, and the highly effective for spectral calibrations.
Joint work with Marina Vannucci, Jerome Bennett, Mike McShane Gerry Coté, Sang-Joon Lee, and Frits Ruymgaart.
Speaker: |
Naeem Siddiqi |
Topic: | |
Date: |
Thursday September 27, 2001 |
Time: |
2:00 - 4:00 p.m. |
Location: |
SAS Canada Executive Briefing Centre Union Station Subway. For those who drive, please don't park on any main streets (King, Yonge, Bay or Adelaide); cars will be towed away from 3:00 to 6:00 p.m. |