Sequential regression techniques with application to the individual sprint in track cycling

The research work described in this thesis is concerned with processes comprising a sequenceof stages, where states and actions taken during each stage influence the outcome at the end ofthe process. Statistical analysis of such processes using standard approaches can beproblematic due to the potentially large number of covariates that are influential, especiallytowards the end of the process. Therefore, three alternative statistical techniques of increasingcomplexity were developed. These techniques are all based on a sequential approach, inwhich logistic regression models are developed at consecutive stages. These techniques wereapplied to the individual sprint event in track cycling and all successfully gave insight intobeneficial tactics for each stage of the race.The first technique involves considering for each model only covariates related to the currentand previous stages. As such, a sequence of overlapping models is created. This approachsuccessfully enabled stable and easy to interpret models to be created. However, the jointeffect of applying tactics at different stages of the individual sprint could not be determined.The sequential logistic regression technique overcame this limitation by using the score (thelogistic transformation of the probability of outcome) from the model developed at theprevious stage as a covariate in the succeeding model. As such, all prior information can beincorporated into each model. However this score is estimated with uncertainty, which cancause the model parameter estimates to be biased. Furthermore, the effects of this intrinsicmeasurement error were found to propagate through stages, particularly in terms of therelative importance of prior and current states and actions. The novel third technique thereforecombines the sequential logistic regression approach with measurement error techniques toaccount for error in the score.

Level dc:type.qualificationlevel

Doctoral (Level 8)

Year dc:date.issued

2012

Author dc:creator

• Moffatt, JL

Language dc:language

en

Identifier

oai:salford-repository.worktribe.com:1338129

OAI identifier oai:identifier

oai:salford-repository.worktribe.com:1338129