@PHDTHESIS\{IMM2003-02459,
    author       = "P. F. Christens",
    title        = "Statistical modelling of traffic safety development",
    year         = "2003",
    school       = "Informatics and Mathematical Modelling, Technical University of Denmark, {DTU,} and Danish Transport Research Center",
    address      = "Richard Petersens Plads, Building 321, {DK-}2800 Kgs. Lyngby",
    type         = "",
    note         = "Supervisor: Poul Thyregod",
    url          = "http://www2.compute.dtu.dk/pubdb/pubs/2459-full.html",
    abstract     = "Road safety is a major concern for society and individuals. Although road safety has improved in recent years, the number of road fatalities is still unacceptably high. In 2000, road accidents killed over 40,000 people in the European Union and injured more than 1.7 million. In 2001 in Denmark there were 6861 injury trafficc accidents reported by the police, resulting in 4519 minor injuries, 3946 serious injuries, and 431 fatalities.

The general purpose of the research was to improve the insight into aggregated road safety methodology in Denmark. The aim was to analyse advanced statistical methods, that were designed to study developments over time, including effects of interventions. This aim has been achieved by investigating variations in aggregated Danish traffic accident series and by applying state of the art methodologies to specific case studies.

The thesis comprises an introduction to accident data, and influential factors such as changing traffic volumes and demographic and economic trends. It highlights the limitations in the influential factors data-structure, in particular, their strong covariance and slow development over time.

An important issue in this thesis was to investigate the temporal dependency in the accident series. The thesis shows that the monthly observations of accidents are serially correlated and that this correlation can only partly be explained by the explanatory variables. One should therefore use dynamic modelling techniques to analyse variations in accident series. The thesis demonstrates that the general decreasing tendency in the accident series has its own slow pattern, not explicable by recorded descriptive variables.

In addition, as a result of the research projects carried out during the preparation of this thesis, I have published the following papers:

- Statistisk modellering af trafik uheld, Trafikdage p{\aa} {\AA}lborg Univeristet, 2001. 
- Sociale karakteristika hos trafikofre, Danish Transport Research Institute, 2001.   
- Models for traffic accidents, {FERSI} Young Researchers' Seminar, 2001. 
- Evaluation of the Danish Automatic Mobile Speed Camera Project Experiment:
 A State-space intervention analysis, symposium i anvendt statistik, 2003.

In Danish:

Trafiksikkerhed er et stort problem, som vedr{\o}rer samfundet og det enkelte individ. Selvom trafiksikkerhed er gradvis forbedret gennem de seneste {\aa}r, er antallet af dr{\ae}bte i trafikken stadig uacceptabelt h{\o}jt. I 2000 dr{\ae}bte trafikuheld over 40.000 i {EU} og skadede over 1.7 millioner. I Danmark i 2001 var der 6861 politirapporteret trafikuheld med tilskadekomst. De resulterede i 4519 lettere tilskadekomne, 3946 alvorligt tilskadekomne og 431 dr{\ae}bte. 

Det generelle form{\aa}l med dette forskningsarbejde er at forbedre indsigten i trafik-sikkerhedsarbejdet p{\aa} aggregeret niveau. Form{\aa}let er at analysere advancerede statistiske metoder, som er udviklet til at analysere udvikling over tid inklusiv bestemmelse af interventioner. Dette form{\aa}l er opfyldt ved at unders{\o}ge variationer i tidsr{\ae}kker af aggregeret danske trafikuheld og ved at anvende state of the art metoder til bestemmelse af specifikke tiltag til forbedring af trafiksikkerheden. 

N{\ae}rv{\ae}rende afhandling begynder med en introduktion af uheldsdata, betydningsfulde faktorer som varierende trafik volume og demografiske og {\o}konomiske tendenser, og fremh{\ae}ver begr{\ae}nsninger i deres datastruktur. Navnlig deres st{\ae}rke kovarians og langsomme udvikling over tid. Et vigtigt emne i denne afhandling er unders{\o}gelse af den tidslige afh{\ae}ngighed i observationer af trafikuheld. Afhandlingen p{\aa}viser, at antallet af uheld er korreleret over tid, og at denne korrelation kun delvis kan forklares ved hj{\ae}lp af de forklarende variable. Derfor b{\o}r man anvende dynamiske modelleringsmetoder til at analysere variationer i tidsr{\ae}kker af trafikuheld. Denne afhandling demonstrerer desuden, at den generelle aftagende tendens i antallet af trafikuheld har sin egen langsomme udvikling, som ikke kan beskrives ved hj{\ae}lp af registreret descriptive variable."
}