Skip to the content

Deterministic evolution of strength in multiple comparisons models : Who is the greatest golfer?

Baker, RD and McHale, IG 2015, 'Deterministic evolution of strength in multiple comparisons models : Who is the greatest golfer?' , Scandinavian Journal of Statistics, 42 (1) , pp. 180-196.

[img] PDF - Published Version
Restricted to Repository staff only

Download (537kB) | Request a copy

Abstract

We present a statistical methodology for fitting time varying rankings, by estimating the strength parameters of the Plackett–Luce multiple comparisons model at regularly spaced times for each ranked item. We use the little-known method of barycentric rational interpolation to interpolate between the strength parameters so that a competitor’s strength can be evaluated at any time. We chose the time-varying strengths to evolve deterministically rather than stochastically, a preference that we reason often has merit. There are many statistical and computational problems to overcome on fitting anything beyond ‘toy’ data sets. The methodological innovations here include a method for maximizing a likelihood function for many parameters, approximations for modelling tied data and an approach to the elimination of secular drift of the estimated ‘strengths’. The methodology has obvious applications to fields such as marketing, although we demonstrate our approach by analysing a large data set of golf tournament results, in search of an answer to the question ‘who is the greatest golfer of all time?’

Item Type: Article
Themes: Built and Human Environment
Schools: Schools > Salford Business School > Business and Management Research Centre
Journal or Publication Title: Scandinavian Journal of Statistics
Publisher: Wiley
Refereed: Yes
ISSN: 1467-9469
Related URLs:
Funders: Non funded research
Depositing User: Prof Rose Dawn Baker
Date Deposited: 30 Jun 2015 18:01
Last Modified: 29 Oct 2015 00:46
URI: http://usir.salford.ac.uk/id/eprint/35659

Actions (login required)

Edit record (repository staff only) Edit record (repository staff only)

Downloads

Downloads per month over past year