Resumen
The monthly road traffic accident victim numbers in Belgium (2003-2014) were analyzed in latent trend time series models separately for pedestrians, cyclists, moped riders, car occupants and road user types jointly. For each road-user type the effect of a range of meteorological variables was tested. The resulting models allow a detailed view on the weather effects for different modes of transport. The strongest effects are observed for two-wheelers (motorcyclists and cyclists), with snow leading to a reduction in victim numbers while warm and sunny weather leads to an increase. The effect of rain differs according to the road user type involved. The principles of state-space time series modelling are described along with the treatment of multicollinearity in models with several predicting variables. An outlook is given of the potential uses of the resulting models.