search

biomodhub / biomod2

BIOMOD is a computer platform for ensemble forecasting of species distributions, enabling the treatment of a range of methodological uncertainties in models and the examination of species-environment relationships.
87 stars 22 forks source link

Variables used in GLM runs #73

Closed luispedrosantiago closed 1 year ago

luispedrosantiago commented 2 years ago

Hi Maya,

I have got a set of presence points for this deer species (n = 126) and am running GLM to produce distribution maps. To do this, I have got a total of 9 raster variables and the number of runs was 100. It turns out when I run the modelling, in nearly all the runs, the selected formula includes one particular variable (which sometimes is the only variable in the formula, really) - WPDA_AmazonBiome_bin, which corresponds to PA in the rasterstack object. On the other hand, some of the variables are seldom picked (or not even picked at all), which results in variable importance values being near 0 (or actually 0), and do not make sense as I know they are biologically important.

I am attaching the code I used and corresponding output (a part of it due to the high number of runs). On the bottom of the attachment, I added the output of the variable importance.

Is this selection made exclusively based on deltaAIC??

Thank you very much. Luís Biomod2.txt

MayaGueguen commented 2 years ago

Hello Luis,

When you say "PA", do you mean "Pseudo-Absences" ? If that is, why do you put it as an explanatory variable ? It is a bit like trying to model the presence of a species and giving the absence points as a variable, it is normal that only this is selected. You should remove it.

Maya