Shrinkage estimation in geographically weighted regression with applications to digital platform pricing
Küçük Resim Yok
Tarih
2026
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Elsevier Sci Ltd
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
Geographically Weighted Regression (GWR) is a spatial statistical technique used to examine how the effects of predictor variables on a response variable vary across different regions. This paper introduces a novel shrinkage estimation approach to improve parameter estimation in spatial regression models, particularly in data-rich environments such as digital platforms and fintech applications. The method enhances predictive accuracy by decomposing the regression coefficient vector into main and nuisance components, with the latter assumed to be close to zero. A shrinkage factor is then applied to adjust the full model estimates towards a more parsimonious submodel, enabling more robust and interpretable results. We provide theoretical justification for the proposed estimators, establishing their superiority over traditional GWR estimators, and demonstrate their effectiveness through extensive geostatistical simulations. Additionally, we apply the method to a real-world dataset from Airbnb pricing in Toronto, showing how the shrinkage approach outperforms conventional models in predicting regional price dynamics and financial outcomes in digital platform pricing settings.
Açıklama
Anahtar Kelimeler
Shrinkage estimation, Geographically Weighted Regression, Digital platform pricing, Spatial financial data, Financial technology applications
Kaynak
Spatial Statistics
WoS Q Değeri
Q1
Scopus Q Değeri
Q1
Cilt
73
