Spatial statistics for real estate data
Kulczycki, Marek
The paper presents spatial statistics tools in application to real estate data, including geostatistics, spatial autoregressive models and geographically weighted regression. All approaches, mentioned above, have different principles but complement each other. Classic statistical methods often fail while having at hand autocorrelated or heteroscedastic data which are natural for real estate. For a long time, spatial autocorrelation or spatial heterogeneity were not taken into account. Last 10 years brought a great interest in employing spatial statistics methods to data spatial in nature such as real estate data what is partially caused by wildly developing GIS software. The content of the paper includes geostatistical methods for localizing real estate submarkets (kriging interpolation) homogenous in respect of price, direct modeling of variance covariance matrix later used in GLS estimation. The application of GWR (Geographically Weighted Regression) for spatial heterogeneity modeling and utilization of spatial autoregressive models for real estate data can be also found. These quite new techniques in authorss opinion, give new opportunities in a field of real estate valuation both by localizing real estate submarkets, their analysis and finally appraisal process. It can be a good tool in mass appraisal (finding taxation zones), in prediction of results of different kind of activities connected with changing spatial planning and also as a supporting tool for decision making process concerning localization of investments.
Event: XXX FIG Working Week and General Assembly : Strategic Integration of Surveying Services
Only personal, non-commercial use of this document is allowed.