Beginning Distance for Incremental Spatial Autocorrelation

12-22-2021 03:13 PM
New Contributor II

Hello, I am conducting a hotspot analysis of disease prevalence rates using census tracts in ArcGIS Pro. I am following the "5-Step Geoprocessing Approach" outlined in this paper (this should jump you to the methods section, you then need to scroll down a little bit to see step 1 etc.):

The issue I am having is with Step 3 where they use Incremental Spatial Autocorrelation  to find the optimal distance band.

I found this presentation from the Michigan State Health Department that cites this very paper and follows the same 5 step process, applying them to preterm birth (PTB) rates as well as low birth weight (LBW) rates by census tract :

This slide from page 11 that talks about Step 3 of the geoprocessing approach (I am going to include the step 2 slide from page 10 as well for context, see below).

I do not understand how they ended up finding distances at which clustering peaks of 4.2 miles and 5.3 miles for their two rates (see the last two lines of the second slide) if the Beginning Distance  for the Incremental Spatial Autocorrelation was set to 20.7 miles? Surely that means that the distances at which clustering peaks will have to be higher than 20.7 miles since in Incremental Spatial Autocorrelation you increment up from the Beginning Distance, not down? What am I missing here?






0 Kudos
2 Replies
New Contributor III

They're using 20.7 miles as the "baseline distance" which I take to mean they compare the Moran's I values of other distances starting at zero and incrementing (as described in the fourth bullet) to the Moran's I value at that baseline distance in order to determine significance at those incremental distances. 

New Contributor II

But the paper in the first link says the following for Step 3:

"We performed the Moran I test at multiple distances using the results from step 2. Our shortest distance tested was two thirds of the maximum distance it would take to reach the 2 nearest neighboring census tracts."

In the slides the " two thirds of the maximum distance" is 20.7 miles, the paper's wording makes it seem like that should be the beginning/shortest distance tested?

0 Kudos