n-1 cross validation method in ENMeval

[arjun nepal]

Good Morning Guys !

I want to ask whether it's possible to use the n-1 cross validation method  while running ENM eval. I have less occurrence points for one species so thought of using that approach but your suggestions would be much appreciated. If it's possible I would love to have the technique for it. 

Another question is I am planning to compare the model of 2 species of the same genus with the niche overlay module of ENM eval. So, is it possible to carry out those comparisons, if I use n-1 CR for one species and random Kfold for another? 

I would love to hear from you all.

Thank You in Advance 

Arjun Nepal (Mr.)

Adjunct Faculty

Royal University of Bhutan 

College of Natural Resources, Punakha

Mobile no: +975_17553882

Alternative Email: nepaala...@yahoo.com

[Ione Arbilla]

Good morning Arjun,

From what I know, you can select the jackknife method in ENMeval, which is kind of similar to the n-1 cross-validation method. I think that the main difference is that jackknife computes statistics from the kept samples only and leaves out the leave-out sample. I don't think there is an option to do n-1 in ENMeval, but maybe someone else can help you out.


Cheers, 


Ione 

[Jamie M. Kass]

Arjun,

Ione is right -- ENMeval implements the "jackknife" method, which withholds one occurrence record for validation and trains the model on the remaining records, then repeats until all records have been validated.

Regarding comparing two species' predictions that had models selected with different cross-validation procedures, this is completely fine. The evaluation method is supposed to be appropriate for the species' data (e.g. if low-data, use jackknife), and this should lead you to select model settings appropriate for that species. When you do a niche comparison, you are comparing the predictions of optimal models for each species, and the way you selected these optimal models should be species-specific.

Hope this helps,

Jamie

[arjun nepal]

Hii Jamie 

Thank you so much for the informations.

Sending you the best regards 

--
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/0b088a74-041b-41df-9a33-d0d6a48a22a1n%40googlegroups.com.