Andreas Dietzel

icon indicating a website link https://www.eawag.ch/de/ueberuns/portraet/organisation/mitarbeitende/profile/andreas-dietzel/show

icon company Swiss Federal Institute of Aquatic Science and Technology (EAWAG) icon location Zurich, Switzerland icon location Quantitative ecologist with background in coral reef ecology, now researching the benefits of Blue Green Infrastructure for urban biodiversity

Results 8 comments of Andreas Dietzel

Varsel stat and stratified cross-validation for highly imbalanced data

Thanks a lot for your answer, Frank. This helps a lot. And apologies, I should have been more explicit in my question or problem statement. I am interested in an...

Varsel stat and stratified cross-validation for highly imbalanced data

Thanks so much for looking further into this. The AUC implemented in projpred does indeed seem to be the AUROC. Implementing additional evaluation metrics is likely currently not one of...

Varsel stat and stratified cross-validation for highly imbalanced data

Excellent. Thanks!

Varsel stat and stratified cross-validation for highly imbalanced data

A short update on this feature request: I have tried to write a new function to calculate the true skill statistic, which is also known as the Peirce score or...

Varsel stat and stratified cross-validation for highly imbalanced data

Thanks a lot for the suggestion, Frank. I haven't had the chance to look into this yet. I have been busy trying to implement cv_varsel() with stratified kfold cross-validation with...

Varsel stat and stratified cross-validation for highly imbalanced data

As Frank pointed out in #352, the bernoulli() distribution can be implemented by using brms:::get_refmodel.brmsfit() instead of init_refmodel().

Make vignette on modelling presence-only data

Excellent package and vignette! I have been working with presence-only data myself and learnt that the AUC is sensitive to class imbalance, particularly when the minority class is of interest....

Parallel rcpp code

If I may hijack this issue on speeding up lsm calculations, I was wondering whether a) any implementation of parallel computing was made in the landscapemetrics package in the end?...