'Final model' when using 10-fold cross-validation

[Jordan M]

Hey everyone,

I'm a new grad student and the first person in my lab group to use Maxent so I'm very new and don't have much help available to me at the moment... Also new to this group!

So I have lots of questions but after doing some practice runs yesterday my main question is about cross-validation:

If I want to do 10-fold cross-validation with 25% test percentage, how do I input that in settings? I did the 25% but when I did 10 replicates and left the replicated run type as cross-validation I got some error about not being able to do cross-validation and also have a test percentage, so I changed cross-validation to subsampling. Is that right at all?

And then once I have my 10 different outputs (so 10 different html files, 10 .asc files etc), how do I get a 'final model'? Do I have to somehow merge all map layers together to get one map layer? Do I average out all the AUC values etc? If I have 10 different html files, which jackknife info do I go by? Or do I use the info from all 10 outputs to change my settings etc and eliminate variables and then run it just once again?

Hopefully those questions make sense and someone can help me out! 

Thanks!

Jordan

[Mokrane Karar]

Hi Jordan, i hope this can help you :    https://groups.google.com/forum/#!topic/maxent/YqVzTIWYi1o

[kaky]

Dear Jordan

As i understand from your question is you used both 10 replication with crossvalidation and 25% test percentage, you can't use both of them together in one setting if you used maxent will give you error and if you press OK automatically the model run and maxent use will choose crossvalidation instead 25%, if you want to choose 25% test percentage you should set the replication for 0. In term of final model there are average for each model in the maxent output you can used as mean for you 10 replications for each species if you use more than one species. I hope this make sense.

Cheers 

Emd