Can UTM data for casualties and obstacles (or in general other structural elements of roads such as bends), be used as distances in spatial autocorrelation analysis?
Very interesting question. From my point of view, you can do it. In ecology, procedures as used for measure the Spatial Auto Correlation (SAC) can be used in order to investigate the congruence between the spatial pattern showed by two variables. Mantel or Moran I test can be used (see Legendre and Gauthier, 2014 for a detailed explanation). The Mantel statistic provide a measure of the correlation between the two matrices and results from the cross-product of the matrix elements (your variables). In fact, the statistic rM is bounded between −1 and +1 and behaves like a correlation coefficient (Fortin and Payette, 2002). Then, the SAC test applied to two variable spatial pattern evaluates the similarity between these two matrices distance (difference in values among neighbour sites) calculated as a geometric distance matrix (see a good explanation in Legendre and Legendre, 2012).
However, I suggest also to take into account some limitations of the method, as highlighted for the case of Mantel test in study SAC: see Guillot, G., Rousset, F., 2013. Dismantling the Mantel tests. Methods Ecol. Evol. 4, 336–344
Thank you very much, Federico! Very useful answer! I have precisely being considering Mantel test, but I have had problems with the input matrices, which are UTM plus other landscape/road features. Thanks very much for the reference suggestions, as well.
I think yes. You should use only the position of the ostacles along the roads and find hot spots. You can use ArcGIS spatial statistics tools and use local Moran or Getis indexes. Please, refer to this articles:
In the new version of ArcMap (10.5 and 10.6) and in ArcGIS Pro a new tool is included in the Spatia Statistics Toolbox "Similarity Search". You ca find how it works in the Online Help of the program.