leave-one-out

81 views
Skip to first unread message

Maja

unread,
Sep 22, 2015, 4:16:09 PM9/22/15
to Maxent

Dear all,


I work with a group of plants having a (very) narrow distribution. For each species I have only a limited number of data of occurrence. I would like to run Maxent using the »leave one out« approach suggested by Pearson & al (2007). I'm a little bit confused how to do it.


Do I have to prepare n datasets with n-1 data of occurrence and run it in Maxent? And, if it is so, what is then my final model? Is the one selected after model selection or the average model, as I could notice in some works?


Pearson & al propose to calculate P using the software accompanying the paper, but as I could notice, this software is not available anymore. Is correct then to use P values proposed for LPT and T10 by Maxent, which I noticed were suggested also by previous posts?


Please, please, illuminate me. Thanks a lot.


Maja

Milton Bastidas

unread,
Sep 22, 2015, 11:14:19 PM9/22/15
to max...@googlegroups.com
Hi Maja, i consider  this paper a good guide:

Shcheglovitova, M., & Anderson, R. P. (2013). Estimating optimal complexity for ecological niche models: A jackknife approach for species with small sample sizes. Ecological Modelling, 269, 9–17. doi:10.1016/j.ecolmodel.2013.08.011


--
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.



--
Milton Enrique Bastidas M.
Licenciado en Biología, UPN
Estudiante de Maestría en Ciencias-Biología
Linea de Manejo y conservación de vida silvestre
Universidad Nacional de Colombia

Jamie M. Kass

unread,
Sep 25, 2015, 11:29:18 PM9/25/15
to Maxent
If you can get your data into R, check out the package ENMeval. You can run multiple models using the jackknife approach and compare them all in one line of code. In brief though, to do the jackknife partition approach, you remove one occurrence pt from your training dataset, run the model, then evaluate your model on that one left out pt. Then you do the same thing for all your pts. Usually then, you average the AUCs and omission rates of all the runs to get an average test AUC and average OR. This is tedious without bringing your data into R, which will go a long way for you. Try out ENMeval -- it's the easiest way to do this.

Jamie Kass
PhD student
City College, NYC

Maja

unread,
Sep 28, 2015, 4:37:43 AM9/28/15
to Maxent

Thank you very much for your help. Regards, Maja
Reply all
Reply to author
Forward
0 new messages