I am interested in applying this methodology to census block data after conversion from a raster:


Basically, the steps are:

  1. Calculate the average distance between polygon centroids - Average Nearest Neighbor tool in Spatial Analyst
  2. Convert Features to Raster
  3. Smooth the outputting raster data by applying the focal statistics tool using the radius calculated in the Average Nearest Neighbor tool.

It seems that, although this may make sense intuitively, because sharp transitions in population count or housing units don't always occur, they may occur sometimes. Though the result is 'nicer to look at', in that it's more a heat map than a rigid rasterized layer, is this workflow statistically defensible? What tests (other than the simple one provided in the article) could you run to test out the smoothed result?

  • If you would like to smooth a discrete density grid obtained from Census block data in a way that respects polygon boundaries, there are simpler and (arguably) better solutions. For instance, smooth the density any way you like (such as a focal mean). Divide the results by the zonal sums of the smooth. Multiply by the block counts. Divide by the block areas. Done. (You will recognize this as being the same idea in my answer to your preceding question, with the smoothed values replacing the relative intensities given by land cover classes.) – whuber Apr 10 '15 at 21:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.