TSS, AUC and other accuracy metric in MaxEnt

[Dimitris Poursanidis]

Hi folks,

The header of the topic is quite familiar to most of us.

My issue is related to the case that we run MaxEnt with a n-fold runs (10-20 runs with crossvalidation,bootstrap, subsampling) and we want to evaluate our results with additional statistics appart from ROC/AUC.

Can TSS be used in a 10-fold cross validation model ? If yes, how ? How you select the logistic threashold as is different in each run ?

D.

[Arta]

Hi Dimitris,

Use the instruction in the forums below (alaaeldeen80, date: 9/29/12)

https://groups.google.com/forum/?utm_source=digest&utm_medium=email/#!searchin/maxent/tss%7Csort:relevance/maxent/eCgJ_0vdOb0/dOLPBkM8KX8J

For more than 1 run you should determine TSS for each and then report the average of them.

Hope it helps.