An analysis of approaches to presence only data

Background/Question/Methods and Results/Conclusions

A number of approaches have been proposed for estimating species distribution based on presence-background data. These include naive logistic regression (treating the background as absence), latent variable models using the EM to unmix the background data, MAXENT, Manly's log-linear model, and other more complex models based on point processes. In this talk I will attempt to clarify the issues, as well as the relationship between these methods. In addition, I hope to propose some new approaches toward these problems.