Examining the effect of daylight on road accidents and investigating a state space time series approach to modelling zero inflated count data

Abstract

In this thesis two aspects of the modelling of road accident count data are investigated in detail.  Under the first investigation the effect of daylight on road accidents is considered.  Here, daylight is established as a significant cause  of car occupant causalities in both Scotland and Southwest England using linear and log-linear regression models.  It is also shown that there is a noticeable difference in the level of daylight during morning rush hour in December and January between Scotland and Southwest England due to the difference in latitude between two regions.  Ad hoc methodology is then introduced to investigate the possibility that the difference in the level of daylight during morning rush hour will result in a significant difference in the numbers of car occupant causalities between the two regions during December and January.

The second investigation considers the use of a conditional Bernoulli truncated Poisson state space time series model for modelling zero inflated count data.  Although it is technically complex, its appeal is likely to be broader than the daylight investigation as the methods presented here offer useful insight into the modelling of any zero inflated time series count data.  The conditional Bernoulli truncated Poisson model has been used before to model zero inflated data, but it has not been used on time series data and has not been put into state space form.  Difficult issues are raised by applying the conditional model to time series data and various methods are introduced and compared to overcome these problems.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata