Szczegóły publikacji
Opis bibliograficzny
Application of robust estimation methods in real estate valuation / Tomasz ADAMCZYK // Acta Scientiarum Polonorum. Administratio Locorum ; ISSN 1644-0749. — 2024 — vol. 23 no. 4, s. 349–360. — Bibliogr. s. 358–360, Abstr.
Autor
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 157437 |
|---|---|
| Data dodania do BaDAP | 2025-01-08 |
| Tekst źródłowy | URL |
| DOI | 10.31648/aspal.10726 |
| Rok publikacji | 2024 |
| Typ publikacji | artykuł w czasopiśmie |
| Otwarty dostęp |  |
| Creative Commons | |
| Czasopismo/seria | Acta Scientiarum Polonorum, Administratio Locorum |
Abstract
Motives: Conducting real estate appraisals in well-developed markets presents a multitude of data analysis challenges. Some property price data may contain outliers that can significantly affect the valuation process and, as a result, the estimated value. In the case of real estate valuation regression models, estimation is most often based on the least squares method, where outliers are taken into account just like the rest of the data. Eliminating or minimizing the influence of outliers can lead to more reliable estimation results. Such problems can be solved by implementing robust regression methods. Aim: The main goal is to determine whether robust estimation methods using M-estimators with multiple regression models can provide more accurate estimates of property values. Results: Research has shown that multiple regression models using robust regression methods can be applied to estimate property values. The use of different types of M-estimators allows for the objective elimination of outliers through algorithms that operate on the entire data set. The calculations are carried out iteratively, and at each iteration step the residuals are verified and the observations are re-weighted. The following M-estimators were considered: Huber, Hampel, Tukey, Faire, Cauchy and Welsch. The reference point was the estimation results from the ordinary least squares method (OLS). All analysed M-estimators led to an increase in the coefficient of determination value and a decrease in standard estimation errors. Each algorithm detected the outliers. The valuation results for the selected properties were also more reliable. The results obtained depend on the characteristics of the data, and the choice of the best estimator may vary across different property markets. The selection of the best estimator may even vary within the same local property market, where the valuer makes subjective assessments of the location or other attributes.
