Current browse context:
stat.ME
Change to browse by:
References & Citations
Statistics > Methodology
Title: Maximum Likelihood Estimation of Spatially Varying Coefficient Models for Large Data with an Application to Real Estate Price Prediction
(Submitted on 22 Jan 2020 (v1), revised 31 Jul 2020 (this version, v4), latest version 12 Nov 2020 (v5))
Abstract: In regression models for spatial data, it is often assumed that the marginal effects of covariates on the response are constant over space. In practice, this assumption might often be questionable. In this article, we show how a Gaussian process-based spatially varying coefficient (SVC) model can be estimated using maximum likelihood estimation (MLE). In addition, we present an approach that scales to large data by applying covariance tapering. We compare our methodology to existing methods such as a Bayesian approach using the stochastic partial differential equation (SPDE) link, geographically weighted regression (GWR), and eigenvector spatial filtering (ESF) in both a simulation study and an application where the goal is to predict prices of real estate apartments in Switzerland. The results from both the simulation study and application show that the MLE approach results in increased predictive accuracy and more precise estimates. Since we use a model-based approach, we can also provide predictive variances. In contrast to existing model-based approaches, our method scales better to data where both the number of spatial points is large and the number of spatially varying covariates is moderately-sized, e.g., above ten.
Submission history
From: Jakob Dambon [view email][v1] Wed, 22 Jan 2020 15:46:29 GMT (3990kb,D)
[v2] Thu, 23 Jan 2020 11:06:00 GMT (3990kb,D)
[v3] Fri, 24 Jan 2020 07:31:48 GMT (3990kb,D)
[v4] Fri, 31 Jul 2020 14:20:51 GMT (3942kb,D)
[v5] Thu, 12 Nov 2020 16:33:17 GMT (3942kb,D)
Link back to: arXiv, form interface, contact.