Setting use of MaxEnt model in the identity.test function

[FernandoMoroC]

How can I set features and regularization coefficient in MaxEnt model when I use type="mx" in identity.test?

[danlwarren]

You should be able to just pass in a "params" vector just as if you were calling it from enmtools.maxent().

[tinsman]

Hi @danlwarren, I used ENMeval to pick some appropriate maxent parameters for the two species I'd like to compare using the identity test. I settled on different parameters for both species. (e.g. L 4 for Species A, which has fewer points than Species B, for which the model built using LQH 1 performs best).
What parameters would you recommend for the pseudo-replicates in that case? Average the regularization multipliers? Select the most (or least) complex features? Blob them both together, re-run ENMeval, and see what comes out there? I wonder what makes the most biological sense.

[danlwarren]

My own personal preference is to go with the simplest model that works reasonably well, but that's just based on my own intuition that these things tend to emphasize discrimination accuracy more than they should, at the expense of biological plausibility. The idea of running the two together is interesting, though, and might be worth pursuing!