Hey all,
I'm modelling the current and future distribution of some fish species using ensemble modelling in biomod2 (presence-only data). I've already built the models and projected them for current and future conditions, but I would like to know exactly which is the threshold that maximizes TSS for binary projections. I've used the BIOMOD_Projection function with the argument binary.meth = 'TSS' but this binary projection for some species just does not make sense (e.g some species get 1 everywhere the probability of occurrence is > 0). For that reason I would lilke to manually find the threshold but I was not able to use the Find.Optim.Stat funciton with presence-only data. Can anyone help me with this?
Cheers
see the links below ....
http://www2.uaem.mx/r-mirror/web/packages/biomod2/biomod2.pdf
https://cran.microsoft.com/snapshot/2020-09-26/web/packages/biomod2/biomod2.pdf
The True Skill Statistic (TSS) is a measure of the accuracy of a binary classification model, such as a species distribution model. In the context of biomod2, the TSS is calculated based on the observed and predicted presence/absence of a species, and it takes into account both the sensitivity (i.e. the proportion of true positives that are correctly predicted) and the specificity (i.e. the proportion of true negatives that are correctly predicted).
To maximize the TSS for a binary projection of a species distribution model in biomod2, you will need to find the threshold that results in the highest TSS score. This threshold represents the probability of occurrence above which a presence prediction is made, and below which an absence prediction is made.
One way to find this threshold is to use the Find.Optim.Stat function in biomod2 with the "TSS" option. This function allows you to optimize the TSS for a binary projection by searching for the threshold that results in the highest TSS score. You can use the function as follows:
Find.Optim.Stat(observed = observed, predicted = predicted, binary.meth = "TSS")
Where "observed" is a vector of observed presence/absence values (0 = absence, 1 = presence) and "predicted" is a vector of predicted probabilities of occurrence.
It is important to note that the Find.Optim.Stat function can only be used with data that includes both observed presences and absences. If you are working with presence-only data, you will need to use a different approach to optimize the TSS for your binary projection. One option is to use cross-validation to estimate the TSS for a range of different thresholds and select the threshold that results in the highest TSS score.
I hope this helps! Let me know if you have any further questions or if you need more assistance with your analysis.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts