Figure 2
From: Application of multiple statistical tests to enhance mass spectrometry-based biomarker discovery
Diagnostic measures comparison of consensus model to the best model from each of the four statistical approaches. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and percent accuracy are plotted for the best model from each statistical approach. When the five parameters are evaluated collectively, the model with the best diagnostic performance is CART, followed by the Consensus model and Logistic Regression. The T-test and UPGMA models have a lower average diagnostic performance as evident in the greater spread of the values of the diagnostic accuracy measures.