Transforming from logit to probabilities in logistic regression models facilitate the interpretation when terms are additive. But, what about interactions?
How can I interpret probabilities when terms include interactions between continuous predictors, just as the same way as interpreting logit relations?