I was reading some literature on spatial modeling, where it is stressed that in using regression (e.g., the Generalized Linear Model family, comprising Logistic Regression for example) the possible spatial autocorrelation of the dependent variable hampers the use of 'traditional' statistical methods. Also, I saw that in many articles spatial autocorrelation is accounted for by including spatial variables into the model itself.
One of such a 'remedy' is introducing into the Regression model an autocovariate term, whose definition can be found at the following link (page 3492 (http://ecology.sdu.edu.cn/english/docs/research_pub_ja07.pdf). Sometimes it is defined as "the euclidean distance weighted mean of all 8 surrounding cells in the raster".
Now, I am planning to model the presence/absence of a land attribute on the basis of some topographic predictors. In anticipation of the fact that the presence/absence of that attribute can be 'affected' by autocorrelation, I was evaluating to calculating that autocovariate term to use as one of the predictors. The problem is that I do not know how to 'translate' the above description(s) into actual GIS procedure. As far as my dependent variable is concerned, my dependent variable is a point shapefile, whose attribute table record the absence (0) or presence (1) of the land attribute. Each point represents a sampling point. Bonus question: how can I formally test for spatial autocorrelation of the presence/absence of the attribute?
أكثر...
One of such a 'remedy' is introducing into the Regression model an autocovariate term, whose definition can be found at the following link (page 3492 (http://ecology.sdu.edu.cn/english/docs/research_pub_ja07.pdf). Sometimes it is defined as "the euclidean distance weighted mean of all 8 surrounding cells in the raster".
Now, I am planning to model the presence/absence of a land attribute on the basis of some topographic predictors. In anticipation of the fact that the presence/absence of that attribute can be 'affected' by autocorrelation, I was evaluating to calculating that autocovariate term to use as one of the predictors. The problem is that I do not know how to 'translate' the above description(s) into actual GIS procedure. As far as my dependent variable is concerned, my dependent variable is a point shapefile, whose attribute table record the absence (0) or presence (1) of the land attribute. Each point represents a sampling point. Bonus question: how can I formally test for spatial autocorrelation of the presence/absence of the attribute?
أكثر...