Trade-offs associated with computational simplifications for estimating spatial statistical/econometric models[Etapes nécessaires afin de rendre l'informatique plus facile pour l'estimation de modèles spatiaux statistiques et économétriques]

Spatial statistics and spatial econometrics seek to account for, or exploit, redundant information that is latent in virtually all geo-referenced data. When ignored, such redundancies compromise traditional parameter estimators, causing certain ones to be biased, most to be insufficient and inefficient, and selected ones to be inconsistent. The cost of retaining these desirable statistical properties, though, is a sizeable increase in computational requirements, or considerable numerical intensity, in some instances to an extent that precludes the proper data analysis altogether. Over the years various attempts have been made to reduce this computational burden. The primary objective of this paper is to initiate an evaluation of benefits and costs associated with implementing computational simplifications when estimating spatial statistical and spatial econometric models.