Maximum likelihood estimation procedures for parameters of generalized gravity model

This research is aimed at developing algorithms for obtaining Maximum Likelihood (ML) estimates of Generalized Gravity Model parameters, which are computationally more efficient than the ones currently available. The model considered is a very general version of the gravity model, which was first presented by Sen and Soot (1981) and which has been given a sound theoretical grounding by Smith (1987).;The proposed algorithms dramatically improve the computational performance of obtaining ML estimates of gravity model. These improvements should assure the relevance of these estimation procedures in the application of the gravity models in transportation planning efforts.;The key modification that dramatically improves performance is a method known as the Linearized DSF (LDSF) Procedure. Specifically five alternative ML estimation procedures are proposed for the general case of more than one measure of separation. The alternative procedures presented are: A Modified Scoring Procedure, three Modified Gradient Search Procedures which I shall call Procedures Ia, Ib and Procedure II, and application of the Generalized Linear Models (GLIM).