The integration of spatial interaction models using generalised linear modelling

Abstract

The aim is to attempt to comprehend apparently separate spatial interaction models using generalised linear modelling. Wilson's entropy-maximising model is formally presented and formulated, using principles of information minimisation and maximum-likelihood, in section 2. In section 3, generalised linear modelling and basic information on the GLIM commands are introduced. The possibility of improvement in spatial interaction models is explored in section 4. Finally, by treating the problem of calibrating spatial interaction models as a problem in statistical estimation, governed by the stochastic properties assumed for the dependent variable, the similarities among several models in spatial interaction research are made explicit. These spatial interaction models are calibrated using generalised linear modelling, whose facilities are available within the GLIM package. -from English summary