Fig. 2

Receiver Operating Characteristic Curve (ROC) of Three Models. ROC measures the true positive rate and false positive rate at all values of probability. The area under the ROC curve evaluates model classification accuracy; the higher the area, the bigger the disparity between true and false positives, and thus the stronger the model in classifying members. The blue colored curve was for the multiple logistic regression model whose area under the curve was 0.85984. The red colored curve was the backward logistic regression model, which has an area under the curve of 0.86875. The yellow colored curve was for the logistic regression model with average scores in all courses, and its area under the curve was 0.85658