In this dataset I have created the green dots represent coyote sightings while the yellow dots represent rabbit sightings. I am trying to determine if coyotes and rabbits are randomly distributed or if there is a relationship between coyote habitat and the proximity of rabbit habitat.

I created a distance matrix between the two sets of data points, but I'm unclear on how to interpret the results. Is there a coefficient that can be used to judge whether or not the distribution appears to be random?

enter image description here

  • 1
    I am somewhat wondering how you want to connect a simple distance matrix to (non) random distribution?
    – Erik
    Dec 8 '20 at 7:56
  • 2
    The problem seems to be more a conceptual one than a technical one. So what exactely is your research interest, what do you want do find out? Have coyotes and rabbits been in the respective places at the same time? If not, does a distance matrix make sense? You should provide more information to get a clear image what you want to accomplish using the data you have. Just processing them is easy, but it should be done with a clear idea in mind.
    – Babel
    Dec 8 '20 at 8:59

You can use a heatmap renderer for your point layers (in the symbology, change from Single Symbol to Heatmap) to see how/if the patterns overlap. That's probably the easiest and fastest way to get a first impression about spatial distribution.

QGIS heatmap symbology renderer

You could also create a Heatmap as raster layer with Menu Processing / Toolbox / Heatmap (Kernel density estimation). You could than even extract contour lines (see screenshot to see how to do that - even though my menus are in german, it should become clear where to find it):

QGIS create raster Heatmap (Kernel density estimation), extract contours

You can do this twice, for each of your points (rabbits and coyotes) - than you have two rasters and two contour line layer:

enter image description here

You can than compare the two contour layers: QGIS overlay contour lines

Instead of contour lines, you can also create contour polygons: Menu processing / Toolbox / Contour Polygons and set a graduated color to the polygons:

QGIS GDAL Contour Polygons

Or you can use the raster calculator to add the coyote-raster to the rabbit-raster to get areas where both of them are often found:

QGIS raster calculator result

That's how the raster calculator looks like - just add the two rasters as shown in this screenshot and define an output path where the new raster should be saved:

QGIS raster calculator input

A different approach (for visualization only) is to render your points as Point Cluster symbology and set the symbol size to data driven override with the variable @cluster_size as seen on the screenshot: QGIS symbology Point Cluster renderer

These workflows are first steps to get an idea about your data. As mentioned in several comments, reflection about the goal of your work and the explanatory power of your hypothesis are crucial and should be answered before proceding data. Get a conceptual understanding of what you want to do and only than define your workflow.

Hopefully the steps presented here can help on the way to get there and to get a better idea about spatial distribution and possible correlations. However, proceding data in different ways is just a matter of technical skills: there are a lot of tools you could use for your data. But that alone will not help you much. To get meaningful results, you should first have a conceptual understanding before you use any software.

  • Thanks, this is helpful. But what test would I run to see if these results have any correlation between them or if they are random?
    – DanG
    Dec 8 '20 at 18:03
  • Use an overlay like the one proposed with raster calculator to see where the regions of high activity overlap. That's what you finally want, right? Otherwise, make clearer what exactly you want to get from the analyse.
    – Babel
    Dec 8 '20 at 18:14
  • I'm trying to determine if the coyote proximity to rabbits is meaningful or if the same results could be observed by chance. These visualizations help to see where coyote/rabbit interactions take place, but I'm not sure how to further answer the question from there.
    – DanG
    Dec 8 '20 at 21:35
  • 1
    @DanG -- If you don't understand your data you'll not know what to ask from it. -- Even if you find spatial auto-correlation between the rabbit and coyote datasets you won't know why. Is it terrain? Fencing? Major bodies of water? -- How close is close enough to make interaction even possible? Are these rabbit warrens? Coyote dens? -- These are biological organisms so their distribution cannot be random across space. -- You ask if the distributions are "meaningful" but if you can't state meaning in relation to some hypothesis you couldn't understand a meaningful answer anyway.
    – user23715
    Dec 8 '20 at 23:46
  • I completely support the comment by @user23715 - make sure you understand what you want to do. The data in itself has no meaning, it' s we that make data meaningful by asking questions. That's what I meant in my very first comment: be sure to understand what you want to do and have a clear goal - a research question or a task to fulfull.
    – Babel
    Dec 9 '20 at 6:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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