BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//AMSE//Event Calendar//FR
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:event-12761@amse-aixmarseille.fr
DTSTAMP:20260415T230812Z
CREATED:20260415T230812Z
LAST-MODIFIED:20260415T230812Z
STATUS:CONFIRMED
SEQUENCE:0
SUMMARY:big data and econometrics seminar - Ghislain Geniaux
DTSTART:20260113T130000Z
DTEND:20260113T143000Z
DESCRIPTION:Spatially varying coefficient models are widely used in fields 
 such as housing markets\, land use\, ecology\, and seismology\, where captu
 ring spatial heterogeneity is essential. Compared to standard Geographicall
 y Weighted Regression (GWR)\, Multiscale Geographically Weighted Regression
  (MGWR) improves estimation by allowing each covariate to operate at its ow
 n spatial scale. Yet\, MGWR relies on a backfitting algorithm that limits s
 calability to moderate datasets and leaves predictive performance largely u
 nexplored. We propose the Top-Down Scale approach for MGWR (tds_mgwr)\, whi
 ch introduces a structured sequence of decreasing bandwidths within the bac
 kfitting process. This avoids full re-optimization at each step\, substanti
 ally reducing computational costs while improving reliability of the global
  optimum. The resulting algorithm\, tds mgwr\, handles up to 50\,000 observ
 ations and 20 covariates efficiently\, combining speed with accurate estima
 tion and enabling more flexible and accurate modeling of complex spatial pa
 tterns. We also introduce the Adaptive Top-Down Scale approach for MGWR (at
 ds_mgwr )\, which incorporates a gradient boosting-like stage to refine cov
 ariate bandwidths sequentially. This captures multiple spatial scales simul
 taneously\, moving beyond the notion of a single optimal bandwidth. Monte C
 arlo experiments show that tds mgwr achieves fast convergence and high accu
 racy\, while atds mgwr excels in complex multiscale settings. However\, app
 lications to real datasets suggest that predictive gains from MGWR over sin
 gle-scale GWR are often modest\, underlining the need for careful cross-val
 idation and AICc-based validation when adopting multiscale models.\\n\\nCon
 tact: Sullivan Hué : sullivan.hue[at]univ-amu.frMichel Lubrano : michel.lu
 brano[at]univ-amu.fr\n\nPlus d'informations: https://amse-aixmarseille.fr/f
 r/evenements/ghislain-geniaux
LOCATION:Îlot Bernard du Bois - Salle 21\, AMU - AMSE\, 5-9 boulevard Maur
 ice Bourdet\, 13001 Marseille
URL;VALUE=URI:https://amse-aixmarseille.fr/fr/evenements/ghislain-geniaux
CONTACT:Sullivan Hué : sullivan.hue[at]univ-amu.frMichel Lubrano : michel.
 lubrano[at]univ-amu.fr
TRANSP:OPAQUE
END:VEVENT
END:VCALENDAR