本計畫將應用現有的空間統計方法於國土資訊系統資料上,並藉由解決實際資料之問題去發展新的統計方法。國土資訊系統乃結合全國各種具有空間分怖之地理資料,如何從這些資料中取得有效的資訊,幫助政府或相關部門做決策成為一重要課題。文獻上常利用核密度估計法(kernel density estimation),將離散型的空間點過程(spatial point process )轉換為連續型的密度函數(density function),此有助於探討該空間點過程與其它變數之相互關係。變數間的影響可能為非線性的,本計畫將會利用Baddeley et al. (2010) 所發展出來的the nonparametric smoothing estimator來判別該變數需要做何種變數轉換,進而配適一個有母數模型,以利於解釋其他因素對該空間點過程之影響。最後利用the smoothed partial residual diagnostic (Baddeley et al., 2010)來檢驗模式是否配適得宜。 This research is intended to apply spatial statistical approaches to the National Geographic Information System (NGIS) data. Through solving a real data problem, a new statistical method might be developed. The NGIS integrates the geographic information for the whole nation’s geospatial data. How to obtain useful information from the NGIS data to help the government or the stakeholders making decision becomes an important issue. In literatures, the kernel density estimation is a commonly used technique to transform discrete spatial points, called a spatial point process, into a continuous density function. The continuous density function is very helpful to study the relation of the spatial point process and some covariates. The covariate effects might be nonlinear. In this research, the nonparametric smoothing estimator proposed by Baddeley et al. (2010) is used to detect the nonlinearity of the covariate effects. A parametric model could be fitted to describe the influence of the covariates to the spatial point process. At the end, the smoothed partial residual diagnostic (Baddeley et al., 2010) is used for model validation.