|
Activity Number:
|
273
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 14, 2002 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Computing*
|
| Abstract - #301830 |
|
Title:
|
MCMC Stochastic Approximations with Application to Ranking Data for Sports Competition
|
|
Author(s):
|
Ming Gao Gu*+
|
|
Affiliation(s):
|
Chinese University, Hong Kong
|
|
Address:
|
Dept of Statistics, The Chinese University of Hong, New Territory, , , Hong Kong, PRC
|
|
Keywords:
|
Monte Carlo ; Ranking Data ; Horse Racing ; Thurstonian Model ; Stochastic Approximation
|
|
Abstract:
|
Ranking data are common in many fields of social, economical and financial investigation. Companies want to know consumers' order of preference; Social and political leaders want to know which factors are important in election campaigns; Sports organizations want to rank their athletes according to the true abilities. Analyses of ranking data have been developed over the years and many models and estimation methods are well established and documented. See, for example, the monograph edited by Critchlow and Fligner (1993) or Marden (1995). However, the most intuitive model, based on the Normal utilities, proposed by Thurstone (1929) has rarely been used. The difficulty lies in the computational difficulty in maximizing the likelihood function. In this talk we present the MCMC stochastic approximation algorithm (Gu and Kong, 1998; Gu and Zhu 2001) and show how the algorithm can successfully be used for estimation with the Thurstonian model. For the purposes of illustration, we present a data analysis with 1000 races from Hong Kong Jockey Club during the 1998/2000 seasons and show a significant iimprovement in investment returns in a back testing exercise.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |
