|
Abstract:
|
An important component of virtually any model of emergency service operations (an analytic model or optimization model) is the travel time of vehicles, especially of those en route to a call. In this paper, we present a semiparametric regression model for estimating travel times for emergency services using unique data from ambulance operations of the city of Calgary, Canada. The model allows nonlinearity of travel time-distance relationship, heterogeneity of variance, non-normality of the conditional distribution of travel times, the effect of other factors such as time of the day, daily trends, etc. Previous approaches of modeling travel times ignore the unique features of the conditional distribution of travel time, especially those of the conditional variance and excess kurtosis. Ignoring the latter aspects of travel time distribution is shown to lead to significant loss of accuracy.
|
