[R-sig-Geo] Inverse Distance Matrix
breitbach at uni-mainz.de
Tue Aug 31 16:52:18 CEST 2010
to pass general weights to a listw object (function nb2listw from the spdep package) for a later Moran's I Test and I decided to follow the recommendation of using the inverse distances as equivalent. I am still not sure how to correctly calculate the inverse distances. I have often seen that it was done by just applying >> 1 / distances << [one divided by distance for all distances]. However I found another form of calculation in "Applied Spatial Analysis with R" where they used >> 1 / (distance / 1000) << [one divided by (distance divided by 1000 for all distances)]. I suppose there is no difference between the two methods since the relation remains the same. Is there any technical reason for the second solution? Most of such functions expect values between 0 and 1 but in this case the values for the latter solution are greater than the original values, which suggests that it would also work with the original values. Unfortunately I found nothing in the help file of the nb2listw function about the range of accepted or preferred values for the glist argument.
More information about the R-sig-Geo