The aim of the thesis is to set out a methodology that includes in time-series modelling exogenous effects measured by additional variables. This methodology is illustrated by a number of applications relating to transport. In these applications, time is measured in days, months, quarters and semesters (half years). The author aims to take account of exogenous effects which are either transitory or durable lasting and which manifest themselves in the short term. The first part of the thesis deals with time-series modelling. The author provides a typology of timeseries models and place our approach within it. We describe the approach used in ARMA modelling with explanatory variables and then in state space modelling with explanatory variables, paying special attention to structural time-series modelling. The second and third parts bring together two groups of applications. The first group considers traffic datasets, for passengers and for freight, aggregated by mode and by main network type. The second group considers numbers of road injury accidents and casualties, aggregated by main network type. The largest period covered is 1970-2000. Most of the applications address the transitory effects on transport demand and road risk of weather and calendar factors. The first detailed results are provided that demonstrate the significance of weather factor on road safety in France, measured by numbers of injury accidents and fatalities. (Author/publisher)
Abstract