Prognosmodeller för antal dödade i vägtrafiken: en utvärdering av Poissonregression med seriellt korrelerade residualer
2008 (Swedish)Report (Other (popular science, discussion, etc.))Alternative title
Forecasting models of the number of fatalities in road traffic. An evaluation of a Poisson regression model with serially correlated residuals (English)
Probability models play an important role when analysing and forecasting time series of the number of fatalities in road traffic accidents. This report presents a study of a method of analysing time series of count data that also can account for serial correlation. The method is a development of traditional Poisson regression where overdispersion and serial correlation is modelled by a latent process. The aim of the study was partly to test how the method worked in practice, partly to study the importance of considering possible serial correlation in the data. The method was tested both for the total number of killed road users and the number of killed road users divided in the following six categories: car occupants (drivers and passengers in passenger cars, lorries, and busses), motor-cyclists, mopedists, cyclists, pedestrians, and others.
Place, publisher, year, edition, pages
Linköping: VTI., VTI rapport 607 , 2008. , 58 p.
VTI rapport, ISSN 0347-6030
Fatality, Accident rate, Forecast, Mathematical model, Probability, Poisson, Regression analysis, Distribution, Correlation, Evaluation, Driver, Cyclist, Pedestrian
Research subject Road: Traffic safety and accidents, Road: Collation of accident statistics
IdentifiersURN: urn:nbn:se:vti:diva-6466OAI: oai:DiVA.org:vti-6466DiVA: diva2:675345