Skip to main content

Table 11 Performance Of Classification Methods used for Class Prediction Analysis

From: Microrna profiling analysis of differences between the melanoma of young adults and older adults

Performance of the Compound Covariate Predictor Classifier:
Class Sensitivity Specificity PPV NPV
Mel 30 0.8 0.9 0.889 0.818
Mel 60 0.9 0.8 0.818 0.889
Performance of the Diagonal Linear Discriminant Analysis Classifier:
Class Sensitivity Specificity PPV NPV
Mel 30 0.8 0.9 0.889 0.818
Mel 60 0.9 0.8 0.818 0.889
Performance of the 1-Nearest Neighbor Classifier:
Class Sensitivity Specificity PPV NPV
Mel 30 0.8 0.8 0.8 0.8
Mel 60 0.8 0.8 0.8 0.8
Performance of the 3-Nearest Neighbor Classifier:
Class Sensitivity Specificity PPV NPV
Mel 30 0.7 0.9 0.875 0.75
Mel 60 0.9 0.7 0.75 0.875
Performance of the Nearest Centroid Classifier:
Class Sensitivity Specificity PPV NPV
Mel 30 0.8 0.9 0.889 0.818
Mel 60 0.9 0.8 0.818 0.889
Performance of the Support Vector Machine Classifier:
Class Sensitivity Specificity PPV NPV
Mel 30 0 0 0 0
Mel 60 0 0 0 0
Performance of the Bayesian Compound Covariate Classifier:
Class Sensitivity Specificity PPV NPV
Mel 30 0.7 0.5 0.583 0.625
Mel 60 0.5 0.7 0.625 0.583
  1. The performance of classification methods used for class prediction analysis in Table 10 was conducted as follows: the Leave-one-out cross-validation method was used to compute mis-classification rate. Based on 100 random permutations, compound covariate predictor p-value = 0.04, diagonal linear discriminant analysis classifier p-value = 0.04, 1-nearest neighbor classifier p-value = 0.02, 3-nearest neighbors classifier p-value = 0.03, nearest centroid classifier p-value = 0.04, support vector machines classifier p-value = 0.72, Bayesian compound covariate classifier p-value = 0.05. For each classification method and each class: Sensitivity = the probability for a class A sample to be correctly predicted as class A, Specificity = probability for a non class A sample to be correctly predicted as non-A, PPV = probability that a sample predicted as class A actually belongs to class A, NPV = probability that a sample predicted as non class A actually does not belong to class A.
  2. T-values used for the (Bayesian) compound covariate predictor were truncated at abs(t) = 10 level. Equal class prevalence was used in the Bayesian compound covariate predictor. Threshold of predicted probability for a sample being predicted to a class from the Bayesian compound covariate predictor was 0.8. % CV support proportion of the cross-validation loops that contained each MiR in the classifiers. T value = ratio of the estimate divided by the standard error.