Can I partition occurrence records into a training set with 60% of the records and the test set with 40% of the records. Or, we can't write a number more than 30 in Random Test Percentage in settings of MaxEnt software?
Please, can anyone help me?
Hi Taghi,
I am not aware of such a restriction.
I just tested it using an old model of mine incorporating 60% as "Random test percentage" in MaxEnt 3.4.1 (using Bootstrapping as "Replicated run type"), and it worked out quite well.
(I had about 200 occurences of a species with a rather small ecological niche).
Crossvalidate, which I believe is the default replicate run type, doesnt use a "Random test percentage" at all, but MaxEnt will tell you that if you hit "Run".
However, keep in mind to still stay within good sample sizes when restricting your training data. When using rather high random test percentages, I would recommend to do a good number of replicates and use the summarized model (in order to account for the lower training data percentage used per each single run).
Hope this helps a little, Florian
Dear Florian;
My occurrence records are about 1500 and when I use
40% occurrence records as test data and Bootstrap method, I acquire the best result.
Anyway Thanks.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping
- Maps