近年來由於地理資訊系統（GIS）的快速發展發，空間資料分析開始受到重視並在社會科學領域中逐漸扮演重要的角色。雖然一般的統計方法已在傳統資料分析上發展已久，然而它們卻不能有效地說明空間性資料，並且無法充分處理空間相依或空間異質性問題。一般而言，空間資料分析主要有兩個分派:模型導向學派與資料導向學派。本文研究目的在於應用空間統計方法合理且充分地評估房地產價值，研究方法包含地理統計(克利金和共克利金)、地理加權迴歸與空間特徵價格模型等，並且以台中市不動產資料進行實證探究。這項新的研究技術在不動產評價領域中將可提供更好的解析能力，使其在評價過程中或是不動產投資決策時，成為一個更強而有力的分析工具。 / In recent years, spatial data analysis has received significant awareness and played an important role in social science because of the rapid development of Geographic Information System (GIS). Although classic statistical methods are attractive in traditional data analysis, they cannot be executed seriously for spatial data. Standard statistical techniques didn’t sufficiently deal with spatial dependence or spatial heterogeneity issues. Generally, the model-driven method and the data-driven method are mainly the two branches of the spatial data analysis. The purpose of this paper is to apply spatial statistics methods including geostatistical methods (kriging and cokiging), geographically weighted regression, and spatial hedonic price models to real estate analysis. It seems to be completely reasonable and sufficient. The real estate data in Taichung city (Taiwan) is used to carry out our exploration. These techniques give better insight in the field of real estate assessment. They can apply a good instrument in mass appraisal and decision concerning real estate investment.

我國公部門以課徵土地稅、發放土地徵收補償為目的而進行土地大量估價，並以區段價法與路線價法為主要估價方法。由於此二估價方法之評估基礎為平均區段地價，故是否能準確劃分地價區段乃至關重要。然而，地價調查估計規則中有關地價區段之劃分規定，並無具體規範劃分準則與方式，導致地價人員僅能透過參考各項可得之圖表資料，並根據自身認知與前輩經驗，將地價相近、地段相連、情況相同或相似之土地劃為同一區段。因此，地價區段劃分之公平、客觀、準確性往往遭受民眾質疑。 劃分地價區段之目的係為掌握地區特性，故其實屬劃分同質區之概念。惟地區特性乃難以定義或量化之區域因素的空間聚集特性，致使地價人員難以掌握具體劃分準則，並準確劃分地價區段。而過去相關研究指出難以定義或量化之空間因素存於特徵價格模型的殘差之中，遂有分析殘差之空間特性以劃分同質區者。但是，各種劃分方式皆侷限於已知樣本所在位置的空間關係，導致可能出現無法就整體地區劃分同質區，或出現預測樣本不屬於任何同質區的情況。 由於克利金法可依據樣本的空間變異結構特性推估未知空間位置的觀察值，因此本研究以區域化變數理論為基礎，應用克利金法結合地理資訊系統之空間分析功能，進而依殘差之空間特性劃分空間效果同質區。研究結果顯示該同質區可合理呈現空間效果之同質性，應用於大量估價的準確性佳，且不會扭曲地價高低層次。因此，應用克利金法劃分地價區段確實為合理、準確且可行之方式。 / Public assessors evaluate official land value for taxing and compensating by land value district approach and street value approach. Since the basis of these two approaches is land value district, whether public assessors could define it accurately or not is an important issue. However, there are no specific defining criterions in Regulations on Land Value Assessment; public assessors could only refer to concerning information, especially their own subjectivity and experience of senior assessors, to define district in terms of “close land value, connected relation and similar circumstances of lands”. Accordingly, district that defined by public assessors not only the fairness and objectivity, but also the accuracy are quite doubtful. The main purpose of defining land value district is capturing local characteristics; therefore, it’s similar to the task of defining homogeneous area. Nevertheless, local characteristics are agglomerations of spatial effect, which are difficult to define or quantity. Due to the fact that public assessors are unable to get specific defining principles, they cannot define land value district accurately. A few researches indicate that spatial effect is in the residuals of hedonic pricing model, thus, some researches defining homogeneous area according to the spatial distribution of residuals. However, the defining approaches of these researches are all restricted to spatial location of known samples. Hence, it’s possible to fail to segregate different homogeneous area, or fail to take unknown samples into consideration. For the reason that kriging estimation can predict unknown spatial location’s value in basis of spatial variation structure characteristics, this research apply kriging estimation and GIS to define homogeneous area based on Theory of Regionalized Variables. The research concludes that homogeneous area which is defined by kriging can capture homogeneity of spatial effect. Besides, the prediction accuracy is quite well by adding variables of homogeneous area to hedonic pricing model. On the other hand, predicted land values still remain the exact relation between each other. Therefore, applying kriging estimation to define land value district is a reasonable, accurate and feasible method.