The objective of this paper is to evaluate the usefulness of the weigh
ted l(p) norm as a distance measure when axis rotation is included to
obtain the best fit. We compare the weighted l(p) and Euclidean norms
in an empirical study of 18 major country and city road networks. A su
bstantial improvement in accuracy is obtained with the weighted l(p) n
orm for networks with a significant component of rectangular travel. W
e show that the weighted l(p) norm, and by inference other multi-param
eter distance estimators, are justified in practice provided the refer
ence axes are oriented properly to model the directional bias inherent
in the network.