Generalizing Impact Computations for the Autoregressive Spatial Interaction Model

We extend the impact decomposition proposed by LeSage and Thomas-Agnan (2015) in spatial interaction model to a more general framework, where sets of origins destinations can be different, relevant attributes characterizing do not coincide with those destinations. These extensions result three flow data configurations which we study extensively: square, rectangular, noncartesian cases. propose numerical simplifications compute impacts, avoiding inversion large filter matrix. considerably reduce computation time; they also useful for prediction. Furthermore, define local measures intra, origin, destination network effects. Interestingly, these aggregated at different levels analysis. Finally, illustrate our methodology case using remittance flows all over world.