作者: K.P. Overmars , G.H.J. de Koning , A. Veldkamp
DOI: 10.1016/S0304-3800(03)00070-X
关键词:
摘要: Abstract In several land use models statistical methods are being used to analyse spatial data. Land drivers that best describe patterns quantitatively often selected through (logistic) regression analysis. A problem using conventional methods, like regression, in analysis is these assume the data be statistically independent. But, have tendency dependent, a phenomenon known as autocorrelation. Values over distance more similar or less than expected for randomly associated pairs of observations. this paper correlograms Moran’s I autocorrelation set Ecuador. Positive was detected both dependent and independent variables, it shown occurrence highly on aggregation level. The residuals original model also show positive autocorrelation, which indicates standard multiple linear cannot capture all dependency To overcome this, mixed regressive–spatial autoregressive models, incorporate were constructed. These yield without better goodness-of-fit. sound presence spatially data, contrast with not. By part variance explained by neighbouring values. This way interactions captured variables. caused unknown processes such social relations market effects.