An empirical assessment of the entropy maximising family of interaction models

Abstract

The parameters, and consequently the performance level, of statistical models of spatial interaction are particularly dependent on the spatial organisation framework which is used to organise interaction data and the way in which the costs of interaction arc measured. This paper provides an empirical assessment of the influence of these factors on the performance levels of the entropy maximising family of interaction models, the members of which differ according to the extent of their data requirements. The principal results which emerge are as follows. The level of model performance generally improves as the number of constraints increases. However, the need for a constraint on the average cost of an interaction appears to be relevant at only one level of spatial organisation. The effect of increasing the number of zones in the spatial organisation framework from eight to twenty-two is only a marginal deterioration in the level of model performance. The use of straight line distances in the calibration of these models does not seriously affect their performance levels. However, a logarithmic transformation of the actual distances produces a significant decline in these levels. The paper concludes with a discussion of the general implications of these findings for intra-regional level analyses of activity patterns in Ireland