I would like to ask a more theoretical question about the methods in the dismo package in R: why is the resulting area under the curve characteristics (AUC) so different between the evaluate() function and the AUC of the test data in the maxent() object?
To illustrate both the AUC of the maxent model and the evaluate, I added the scripts that I used to calculate these two values.
Maxent object:
## assuming a random occurrence data set(my_occ_data) and environmental layer stack(s)
me <- maxent(x=s, p = my_occ_data, args = 'randomtestpoints=20')
maxentResults <- read.csv("maxentResults.csv")
AUC <- mean(maxentResults$Test.AUC)
Evaluate():
fold <- kfold(my_occ_data, k=5)
occtest <- occ[fold == 1,]
occtrain <- occ[fold != 1,]
me <- maxent(x=s,
p = occtrain)
e <- evaluate(me@models[[1]], p=occtest, a=bg, x=s)
AUC <- e@auc
I would expect these two methods to yield similar results but the first method of using the maxent AUCs results in much lower values for my models....... Many thanks if you could share some knowledge on this.
