[R-sig-Geo] auto-correlation of nominal variables

[Rich Shepard]

On Sun, 16 Jun 2019, Emanuele Barca wrote:

> I received a  dataset of point data, organized in the following way: a
> couple of coordinates and a column of "crop classes codes", three columns.

> I would like to compute something similar to the Moran index for each crop 
> class code.

Emanuele,

What is your interest in these data? Are you interested in the pattern of
crop classes (such as in epidemiology or designing a timber sale) or
something else?

If it is the pattern that is of interest I suggest you first test for
complete spatial randomness. Look at some of the references and R packages
for spatial point processes. The distribution of crop classes could be
random, regular, or clumped/clustered.

Peter Diggle's book, "Statistical Analysis of Spatial Point Patterns," and
"Spatial Point Patterns" by Adrian Baddeley, Ege Rubak, and Rolf Turner
would be excellent places to start.

Hope this helps,

Rich