I have a set of points with attributes populated which will be input of logistic regression model.I want to minimize the effect of spatial auto-correlation by taking sample from these points. I have approached building grids(fractals) by a spacing and taking exact one sample from each of these grids.Now my question is how can I get the dimension of the grids(fractals) for which the auto correlation (MORAN's I) will be least? Or any better idea to minimize spatial auto correlation among the points?