There are often time dependent observations within traffic safety research. This can vary from the number of accidents for a certain road section or area, to the vehicle's speed and position for each second. It appears in practice that there is a large category of time dependent observations that can be formulated as linear dynamic state space models. The model is examined that determines: (1) how the input quantities (circumstances) influence the state space quantities, and (2) how the state space quantities determine the output quantities (result). Interpreting the parameters of this model is emphasized here, mainly in an exploratory sense. This makes it possible to indicate the relative importance of the different input quantities or intervention possibilities for the output. The paper also indicates how the Kalman filtering technique can be applied to multidimensional state space models.
Abstract