Model estimation of the longevity for cars registered in Sweden using survival analysis and Cox proportional hazards model

by Daniel Söderberg

Institution: Uppsala University
Year: 2014
Keywords: Survival analysis; Cox proportional hazards model; Trafikanalys; car longevity; Natural Sciences; Mathematics; Probability Theory and Statistics; Naturvetenskap; Matematik; Sannolikhetsteori och statistik; Masterprogram i statistik; Master Programme in Statistics
Record ID: 1338226
Full text PDF: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-227520


Time-to-event data is used in this thesis to analyze private cars’ longevity in Sweden. Thedataset is provided by Trafikanalys and contains all registered, deregistered or temporary deregisteredcars in Sweden during the time period 2000 - 2012.A Cox proportional hazards model is fitted, including variables such as car manufacturer andcar body. The results show that directly imported cars have a much shorter median survivalcompared to non-imported cars. The convertible cars have the longest median survival amongthe five different car bodies. Sedan and station wagon body types have the shortest mediansurvival. Volvo and Mercedes have the longest survival while Renault, Ford and Opel have theshortest survival. The model fits the data reasonably well, and the assumption of proportionalhazards holds for most of the variables.