Table 4.
Parameters summarizing the performance of each model on the test set.
| Classification performance | ||||||
|---|---|---|---|---|---|---|
| OPLS-DA | ROC (AUC)a | Sensitivity (%)b | Specificity (%)b | Precision (%)b | Accuracy (%)b | CV-ANOVA p valuec |
| Model 1: CN–AD | 0.99 | 93 | 65 | 73 | 79 | 0.003 |
| Model 2: CN–MCI | 0.88 | 88 | 44 | 61 | 66 | 0.01 |
| Model 3: SMD–AD | 0.96 | 95 | 45 | 63 | 70 | 0.001 |
| Model 4: MCI–AD | 0.97 | 92 | 50 | 65 | 72 | 0.008 |
aThe area under the ROC curve (AUC) of the receiver operating characteristic (ROC) curve was calculated for each binary classifier in terms of the associated YPredPS values in predicting the class membership for samples of the test set. YPredPS is the Y value predicted by the model based upon the X block variables (resonance intensities at given ppm). An YPredPS value close to 1 would indicate that the subject is likely to belong to the class. An YPredPS value close to 0 would indicate that the subject is unlikely to belong to the class.
bSensitivity, Specificity, Precision and Accuracy levels were obtained from the constructed Confusion Matrix combining the different fractions of true positive (TP), false positive (FP), true negative (TN) and false negative (FN) values for each classification.
cp value for the OPLS-DA reliability (p < 0.05).