Using Maxent with true presence/absence data

[Heather Welch]

Hi all,

I'm interested in the group's thoughts about running Maxent on a true presence/absence dataset. The samples-with-data format allows the user to give Maxent pre-defined background points (in this case true absences) to build the model, therefor allowing Maxent to make use of true absences. My interest in using Maxent as opposed to a modeling method specifically designed for true presence/absence data is 1. I am familiar and comfortable with Maxent and have working code, and 2. Maxent run with presence only data has produced models comparable to those produced by modeling methods designed for p/a data (e.g. Gormley et al 2011 - Using presence-only and presence-absence data to estimate the current and potential distributions of established invasive species).

My questions to the group are:
1. Is Maxent run on p/a data implicitly inferior to other modeling methods designed for p/a data?
and
2. Can anyone point me towards relevant literature regarding Maxent and p/a data?

Cheers,
Heather

[Bill Sutton]

Heather, 

Before I give my opinion, can you provide a little bit of information on what other types of modeling techniques you were considering along with the study organism(s) you are modeling? It's likely to affect the group's response somewhat.

Thanks!

Sent from my iPhone

--
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.

[Jamie M. Kass]

Most of the literature recommends the use of p/a techniques for p/a data, which includes the suite of regression- and tree-based methods (glm, gam, brt, random forest, etc.). If you have multiple site visits, you should certainly go the occupancy model route. One thought - if your absences cover an environmentally unique subset of your study extent, a background approach would likely give you an underpredictive model. But please consult the literature first.

Jamie Kass
PhD Student
City College, NY

[Francisco Rodriguez Sanchez]

Hi Heather,

This recent paper looks relevant to your question: http://onlinelibrary.wiley.com/doi/10.1111/2041-210X.12252/full. In brief, when you have P/A data
the authors recommend using P/A methods rather than Maxent.

Cheers,

Paco

--
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.

-- 
Dr Francisco Rodriguez-Sanchez
Integrative Ecology Group
Estacion Biologica de Doñana - CSIC
Avda. Americo Vespucio s/n
41092 Sevilla (Spain)
http://bit.ly/frod_san