Background for the Model:
Data I have:
- Forest loss points
- Forest boundary
- Randomly selected non loss points
(1 & 3 are response variable)
What I need: I need to minimize the effect of spatial autocorrelation in the regression. For this purpose I need to take points while there is least autocorrelation among the taken points, I think.For this purpose I need to take grids, I think, represent autocorrelation among grid-inside points and least among the points of different grids.So now I need creating those grids for which I need to know the dimension of the grids.
How can I create those grids.
What I have tried so far: I generated square grids at different spacings and spatially aggregated points by those grids. Finally I ran incremental spatial autocorrelation among those grids at different increment steps but this does not give me a peak point what I would take as grid dimension. Another problem this process creates it gives me a peak point at ~25000m but if I take this distance as the grid dimension then the grid number will be too small and thereby the sampled points would be also too small for fitting the logistic regression.