What is the purpose of the "tails" on maxent-generated response curves (plots)

96 views
Skip to first unread message

Cooper Marcus

unread,
Oct 28, 2021, 9:56:07 AM10/28/21
to max...@googlegroups.com
When reviewing covariate response curves generated by maxent, we often see these little "tails"

image.png
These "tails" are also present in the .dat files that contain the data represented in the charts:

image.png

What is the purpose of these tails? Do the data points in the tails tell us anything?

Or are they spurious, have no meaning, and can be safely removed?

Thanks, Cooper

Ahmed El-Gabbas

unread,
Oct 28, 2021, 11:06:32 AM10/28/21
to Maxent
Hello Cooper,

I think this is an effect of using clamping. By using clamping, maxent fixes predicted values beyond the training range to zero.

Cheers,
Ahmed

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Ahmed El-Gabbas,
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Cooper Marcus

unread,
Oct 28, 2021, 11:48:57 AM10/28/21
to max...@googlegroups.com
Do the tails express any useful information? How should I interpret them?

Or can they be ignored?

(Asking because we are showing these charts to stakeholders as we teach them how to interpret the model, and the tails have been causing confusion - I'm wondering if I can safely trim them off :)

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/maxent/b2d59708-1d76-4df3-abc0-10b54d5378fan%40googlegroups.com.

Samuel Veloz

unread,
Oct 28, 2021, 4:39:27 PM10/28/21
to max...@googlegroups.com
I think this showing where clamping is occurring. The predicted values are constant as the x variable increases or decreases from where the tails begin. These tails start at the extreme values of the predictor in the training data set. So if you are projecting to new data that has values beyond these extreme values, you are extrapolating beyond the data used to train the model.
Sam

Cooper Marcus

unread,
Oct 28, 2021, 6:01:37 PM10/28/21
to max...@googlegroups.com
Ah, that makes sense, thanks much for your quick and helpful explanation!

Belay Ejigu

unread,
Oct 30, 2021, 9:05:39 AM10/30/21
to max...@googlegroups.com
Dear bro I coudn't understand your question. 

Another from my side dear brother, have got the secret formula or technique with which MaxENT model algorithm calculate the probability of similarity between the environmental variables outside the range of the realized niche of species and that of observed presences?. 

Please I need some clarifications on this. In short, how the maxent model determine the suitability level of each variables outside the range of observed niche?.

Thanks a lot in advance!
Stay safe and blessed ever with your family!


Reply all
Reply to author
Forward
0 new messages