Interpreting maxent output raster values

[David Beasley]

Please pardon my naive question, I am a GIS student with limited statistical background trying to interpret maxent results.

Am I correct in my understanding that the values of the rasters represent a probability distribution such that the value at every pixel indicates the probability of species presence at that pixel?

I'm trying to wrap my head around the varioius threshold values in the maxentresults.csv output.

Thanks,

David

[romunov]

Hi,

you might want to check out

http://web.stanford.edu/~hastie/Papers/maxent_explained.pdf

Cheers,

Roman

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxent+un...@googlegroups.com.
To post to this group, send email to max...@googlegroups.com.
Visit this group at http://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

--

In God we trust, all others bring data.

[David Beasley]

Thanks for that Roman, but you missed the part I wrote about my "limited statistical background." :)

[Jamie M. Kass]

In short, the raw output is interpreted as a "relative occurrence rate", where the value of each pixel is the predicted occurrence rate relative to all other pixels in your study extent. This is NOT the same as an occurrence probability, which is accurately estimated with an occupancy model that estimates both occurrence and detection probabilities. The logistic output is an attempt at rescaling the raw output to be between 0 and 1, and is usually interpreted as "suitability" (or sometimes probability of presence), but includes the big assumption that the probability of presence at "average" locations is 0.5. If this is not the case for your species, you need to think about the implications of this assumption. A good paper to read that covers this well is "A practical guide to MaxEnt for modeling species’ distributions: what it does, and why inputs and settings matter" (Merow et al. 2013). Even if you have a limited statistical background, you'll need to brush up a little if you want to interpret these models correctly. But you can easily skip the formulas in these papers and read the content instead, and you should understand the important stuff.

Jamie Kass
PhD Student
City College, NYC