2
$\begingroup$

I am new to spatial statistics and need help figuring out how to assess if there is spatial clustering in my dataset using R.

I am assessing if acorn consumption is different inside timber harvest gaps than the adjacent unharvested forest. To test this I set up acorns in grids that spanned from inside a timber harvest gap into the unharvested forest. The grids are t-shaped to sample the same area along the harvest gap edge as the gap interior and unharvested forest. I’ve plotted the grids and color-coded the binomial acorn fates (example graph and data subset below, missing points are sham caches where there was no acorn). I want to test if acorn fate is spatially clustered and am struggling to shift through all the options available.

Many of the tools I have looked at calculate if the points themselves are clustered in space, but I want to know if the results are spatially clustered.

It seems like calculating a measure of spatial clustering then using a randomization test to calculate a null distribution for every gap is promising, but I cannot figure out a simple spatial clustering metric to use for this method. I am also open to others this is just one thing I was thinking about.

Further, I would also be interested if the three areas (harvest gap, edge, and unharvested forest) display different degrees of spatial clustering in addition to the degree of spatial clustering relative to random in the grid as a whole.

Below is a plot of one of the grids and a subset of the dataset for that grid. I would like to do this analysis in program R.

Thanks in advance for the thoughts!

enter image description here

enter image description here

$\endgroup$
  • $\begingroup$ Could you explain what the values in the columns mean and how they relate to the information in the map? $\endgroup$ – whuber Jan 12 '18 at 17:33
  • $\begingroup$ 'Loaction' is a factor describing the general location in the grid (basically left side, middle, right); 'Final fate" is a binomial variable for if the acorn survived (1) or was consumed (0); x and y are the grid coordinates used to plot the points. I want to test if there is spatial clustering of the 'final fate' variable. $\endgroup$ – Skye G. Jan 12 '18 at 17:42
  • $\begingroup$ Thank you. But in that case the table you characterize as the "full dataset," which has only nine rows, doesn't seem to match the map at all, which displays far more than nine points. $\endgroup$ – whuber Jan 12 '18 at 17:46
  • 1
    $\begingroup$ I am sorry I thought that I could attach a copy of the dataset, but did't realize that was not possible on stack exchange and forgot to remove that from the original post. There are 75 datapoints for each grid, not 9. I will edit it, thank you for pointing that out! $\endgroup$ – Skye G. Jan 12 '18 at 17:52
  • $\begingroup$ I like your idea of creating a null distribution to test against. I suppose you could use something like Euclidean distance, but that doesn't seem like an ideal approach given that it would lead to distances that essentially traverse areas outside the target grid. Perhaps you could use a different measure based on the proportion of "matches" for each plot in proximal cells? $\endgroup$ – Matt Barstead Jan 12 '18 at 22:59

Your Answer

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

Browse other questions tagged or ask your own question.