# How to test for correlation between two point data sets

I want to determine if there is a correlation between weed plant species density and habitation (buildings).
Both data sets are points. The weed species has an 'abundance value' (i.e., one point may represent any value captured in an attribute field). The habitation point data set is one 'building' per point.
I want to test both; for 1) all species relationship to buildings, and 2) per single species to buildings (i have 115 species). The relationship should be tested up to about 3000 m from each species point. Locations are not coincident.

I only have ArcGIS 10.4 with Spatial Analyst, Geostatistical Analyst, and other common extensions. I would also be able to use QGis.

I have little knowledge of R, python, etc.

My hypothesis would be that 'weed species are more abundant close to habitation and more so the denser habitation is'.

• Thiessen polygons might be a good place to start.. Larger polygons represent lower density (1 per shape_area) or Near/Generate Near table with a maximum distance of 3km. If you want to use raster then Euclidean Distance is a tool that might be useful. Though last time I did an analysis like this I only proved that there were more samples taken in close proximity to roads as the data capture volunteers were less likely to walk a great distance from their vehicle, the natural conclusion was that koalas like asphalt. – Michael Stimson Apr 26 '17 at 1:40
• Derive distances to buildings and summarise distance per abundance class – FelixIP Apr 26 '17 at 1:44
• Thx @MichaelMiles-Stimson and FelixIP. I derived distances already as suggested and then stared at the answer but i'll summarise as suggested by Felix when i get back to work (4 am here) and then it may make sense to me. I'll also do Thiessen polygons, thx, to visualise it better. In the end this goes into an article so i have to 'describe / validate the answer statistically' but think it will be possible after the 'summarising' as suggested. Thx again. I'll look at other posts as well as they come in. – Johan Baard Apr 26 '17 at 2:07
• If you are doing 115 separate statistical tests for your 115 weed species then watch out for multiple testing problems. xkcd.com/882 applies. – Spacedman Apr 26 '17 at 6:59