[R-sig-Geo] analysis of point patterns

[Adrian Baddeley]

Lars-Daniel Weber <Lars-Daniel.Weber at gmx.de> writes:

>I've got a road network (polylines) with points on it as geodata. The
>points have quality measurements of the paving within a distance of 10 to
>20 metres. I'd like to run a analysis on these points to detect
>autocorrelation and hot spots with good and bad paving quality: high-high
>(etc). Local Moran's I might be a good solution from my point of view.
>
>I thought of creating 5 classes of quality from 1 (bad) to 6 (perfect)
>and put it into Local Moran's I. Is this a good approach to get some
>analysis out of the data?
>
>Which packages would you suggest? Or would it be better to write it on my
>own? I can export the point data to LAT;LON;CLASS to do easy processing
>in R.


The R package SSN is designed for geostatistics on a network (originally,
river and stream networks) and has full functionality but is initially
restricted to networks without loops. However it may also be able to
handle a general road network, depending on what you need.

In the R package Œspatstat¹ there is infrastructure for data on linear
networks (classes linnet, lpp etc). The current release does not cover
geostatistical methods, but you can compute shortest-path distances in the
network and construct your own Moran¹s I.


Adrian Baddeley