Skip to main content

Table 4 Comparison of performance of the two datasets (10-fold cross validation [CV] and permutated model), generated using CV test for the best combos for the 3 PD groups

From: Selected serum cytokines and nitric oxide as potential multi-marker biosignature panels for Parkinson disease of varying durations: a case-control study

Duration of PD

Dataset type

ACC

Error rate

SE (%)

SP (%)

AUC

<  1 year PD

10 fold CV

0.809

0.191

83.3

80.4

0.868

Permutated model

0.632

6.177

62.5

66.2

0.615

1–3 years PD

10 fold CV

0.744

0.256

96.7

62.5

0.817

Permutated model

0.626

7.939

62.0

64.7

0.608

3–12 years PD

10 fold CV

0.884

0.116

93.3

85.7

0.904

Permutated model

0.616

7.844

60.4

65.2

0.605

  1. ACC Accuracy, SE sensitivity, SP specificity, AUC area under curve
  2. This compares the performance of our data with a permutated model (where the analysis was repeated N times) with a new data set on the y-axis to estimate new AUC values to provide an estimation on the overall panel performance with higher SE and SP