Binary Logistic Regression Performance Plot and ROC curve

Mike_Szymczuk

For binary logistic regression's performance plot and ROC curve, cut-off steps are may be added at various increments, .2 is the default. What do the .2 step values mean?

Kucharssim

Dear @Mike_Szymczuk,

thank you for your question.

These cutoffs are based on the probability scale, so for instance 0.8 cut-off means that if a data point is predicted to have at least 80% probability of being in the "positive" category, it will be classified as "positive". By default, the analysis uses cut-off of 0.5.

I see that we lack documentation for this plot, I will see to add this explanation.