Prediction capability of Species Distribution Models are generally measured through ‘Area Under Curve (AUC)’ using training data sets. I have developed four distribution models of my study species. All models are in raster format. Training data sets are exact GPS locations of the species. Can anyone help me! How to create AUC using R commands? Which R package can be used? Please share your R codes if available!
Another alternative would be the "PresenceAbsence" package.
If you have a file called "validate_data" with this structure:
ID Observed Predicted
1 0 0.15
2 1 0.86
3 1 0.92
4 0 0.26
Just write:
auc (validate_data) # Get the AUC value. You can specify the sd.
auc.roc.plot (validate_data) # Plot the ROC
There are many packages like ROCR package and many code samples on the internet. Just have a search
package 'dismo' is good too.
https://cran.r-project.org/web/packages/dismo/index.html
Another alternative would be the "PresenceAbsence" package.
If you have a file called "validate_data" with this structure:
ID Observed Predicted
1 0 0.15
2 1 0.86
3 1 0.92
4 0 0.26
Just write:
auc (validate_data) # Get the AUC value. You can specify the sd.
auc.roc.plot (validate_data) # Plot the ROC
See biomod2 function
Find.Optim.Stat(Stat='ROC', Fit=pred_trained, Obs=obs_trained)
See http://rpackages.ianhowson.com/cran/biomod2/man/Find.Optim.Stat.html
Their example looks nice.
See Model fitting, prediction, and evaluation section from this link: http://rspatial.org/sdm/rst/5_sdm_models.html
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package