How to generate ROC curve using classifier model trained on one data set but tested on a new data set?

MATLAB 2015a classification learner app has used features from 91 images to train cubic svm and ensemble bagged trees classifiers using the No Validation setting. 1) What parameters are being adjusted to calculated the ROC curves associated with these classifiers and 2) How can ROC curves be calculated in MATLAB when exporting these two classifier models and testing them on new data sets?