[R-sig-Geo] dismo library: species distribution modelling and obtaining thresholds values

Robert Hijmans r.hijmans at gmail.com
Mon Jun 27 17:28:07 CEST 2011


> I am using the dismo library for species distribution modelling using a 
> number of models. I would like to select different threshold levels. The 
> sdm.pdf vignette only gives an example to obtain the Maximum training 
> (or testing) sensitivity plus specificity threshold: 
> e at t[which.max(e at TPR + e at TNR)] 

> I was wondering how one would go about obtaining the following 
> thresholds from a ModelEvaluation object: 
> Equal training sensitivity and specificity 
> Balance training omission, predicted area and threshold value 
> Equate entropy of thresholded and original distributions 

Alastair, 

There is some need for further development here. But for "Equal training
sensitivity and specificity" you could probably do

e at t[which.min(abs(e at TPR - e at TNR))] 

"threshold value" would be one of the values in (or the nearest to) e at t (t
stands for threshold), and you can set these with an argument to evaluate.

I do not know what "balance training omission" is (balance with what?).
"Predicted area" is also vague for a threshold: relative to what?

See str(e) for what else is in there. 

Hope this helps,
Robert


--
View this message in context: http://r-sig-geo.2731867.n2.nabble.com/dismo-library-species-distribution-modelling-and-obtaining-thresholds-values-tp6509338p6520941.html
Sent from the R-sig-geo mailing list archive at Nabble.com.



More information about the R-sig-Geo mailing list