Variables contribution and mapping of model predictions using presence-only data

[flopesdematos]

Dear Babak,

I'm trying to extract the variables importance and map the model prediction of models fitted using presence-only. However, the following errors pop-up:

• for the getVarImp() function -> Error in o[[1]] : subscript out of bounds
• for the predict() function -> Error in .generateWLP(x = object, newdata = newdata, w = w, species = species, : the data does not contain some or all of the variables that the model needs...

I include a reproducible example:

library(sdm)

file <- system.file("external/po_spatial_points.shp", package="sdm") # path to a shapefile

po <- shapefile(file)
class(po) 

head(po) 

path <- system.file("external", package="sdm") 
lst <- list.files(path=path,pattern='asc$',full.names = T) 

preds <- stack(lst) # making a raster object

d <- sdmData(formula=sp4~., train=po, predictors=preds)

d

m <- sdm(sp4~., data=d, methods=c("bioclim","domain"),
          replication='sub', test.percent=30, n=10)

getModelInfo(m)

getVarImp(m) 

p1 <- predict(m, newdata=preds)

There is an easy fix for this issue?

Thank you!

[babaknaimi]

Hi,

I understand that the methods you selected are presence-only, but it would currently be better to generate background as well, because they would be needed for model evaluations, so adding bg to sdmData would fix the issue.

d <- sdmData(formula=sp4~., train=po, predictors=preds,bg=list(n=100))

Hope this helps, Babak