I'm using the "Average Nearest Neighbor" tool in ArcGIS 10.3 to calculate whether my point feature class is clustered, random or dispersed.

One of the parameters is area. I tried putting different areas and discovered that if I cut the area lets say by 50% the result changes form clustered (one extreme of the graph created by the software) to dispersed (the other extreme of that very same graph).

Can someone please explain how area is being used by ArcGIS to calculate that? As far as I knoe Nearest Neighbor has nothing to do with area it's all about the distance between the points.

  • From the tool notes, area is the size of the study area. If you reduced it by 50%, it may not be using all the data.
    – mkennedy
    Apr 14, 2016 at 16:56
  • that can explain it. Although in terms of size it is absolutely enough. Is there a way to tell it where (which corner) to start the area from? Apr 14, 2016 at 16:59
  • 1
    @mkenned As I recall, the tool will use all the data. The numerical value of the area is used to estimate expected intensities. If you input an large area, the expected intensities will be low and so the actual intensities will seem too high in places, leading to a determination of "clustering." If you input a small area, the expected intensities will be high and the actual intensities will be generally too low, leading to a determination of "dispersed." Unless the study region was determined a priori, this "area" input is arbitrary, making the tool practically worthless--and even deceiving.
    – whuber
    Apr 14, 2016 at 18:36

1 Answer 1


As @whuber states: "Unless the study region was determined a priori, this "area" input is arbitrary, making the tool practically worthless--and even deceiving". This is advice work heeding.

The nearest neighbor index is the ratio of the observed and expected mean neighbor distances. The expected is a function of the area and the number of observations.

expected mean nearest neighbor distance = 0.5 * sqrt(area / n)

Because of this, area can have a profound influence on the resulting statistic. I believe that ESRI put this option in to allow the user to define the area of a custom polygon region. Certainly, it is not intended to test the effects of different arbitrary areas.

It is not best statistical practice to define your study area using the extent or convex hull of your observations, which is the default. For example, if you have a process that is hypothesized to be occurring within a watershed, you would want to use the watershed boundary and not a convex hull of your observations.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.