Hello,
I am looking for help in determining the minimum number of features to calculate reliable indices for Global Moran's I spatial auto-correlation using -spatgsa- command (I use first order polygon contiguity for the spatial weights matrix, -spatwmat-). I am attempting calculations for Moran's I for a total of 922 areas, comprised of anywhere between 3 and 2,871 polygon features each. I am wondering what is the best accepted practice /cut off point for the minimum number of features that can be used in order to get at a reliable value for the index. After having generated all of the indices, it seems that the areas with low numbers of features almost consistently produce a negative value for the index. So just want to find out at what level it would make sense to cut/drop areas with too few features. Thank you for any help.
I am looking for help in determining the minimum number of features to calculate reliable indices for Global Moran's I spatial auto-correlation using -spatgsa- command (I use first order polygon contiguity for the spatial weights matrix, -spatwmat-). I am attempting calculations for Moran's I for a total of 922 areas, comprised of anywhere between 3 and 2,871 polygon features each. I am wondering what is the best accepted practice /cut off point for the minimum number of features that can be used in order to get at a reliable value for the index. After having generated all of the indices, it seems that the areas with low numbers of features almost consistently produce a negative value for the index. So just want to find out at what level it would make sense to cut/drop areas with too few features. Thank you for any help.
