Skip to main content

Table 2 Comparison of methods with different correlation levels (ρ) and signal noise ratio (SNR) from 100 simulation datasets

From: Application of penalized linear regression methods to the selection of environmental enteropathy biomarkers

ρ

SNR

Criteria

Stepwise

Elastic Net

LASSO

MCP

SCAD

Adaptive LASSO

Oracle

 

1

TP

2.30

4.44

4.48

3.78

4.26

3.74

7.00

  

FP

0.30

4.06

4.20

2.30

3.39

2.04

0.00

  

MRME (%)

60.40

46.34

46.56

45.04

43.51

45.23

17.93

0.2

3

TP

4.97

6.51

6.55

6.05

6.35

6.08

7.00

  

FP

0.35

4.75

4.80

2.04

3.34

2.08

0.00

  

MRME (%)

65.26

61.87

62.17

56.00

59.48

54.50

17.93

 

5

TP

6.16

6.89

6.88

6.68

6.76

6.71

7.00

  

FP

0.41

5.30

5.32

1.93

3.21

2.06

0.00

  

MRME (%)

46.40

61.05

61.65

49.76

52.87

47.44

17.93

0.5

1

TP

2.29

4.38

4.41

2.66

4.15

3.67

7.00

  

FP

0.34

4.16

4.11

2.38

3.36

2.17

0.00

  

MRME (%)

57.76

46.01

46.56

44.03

43.00

44.66

17.89

 

3

TP

4.95

6.54

6.54

6.01

6.31

6.00

7.00

  

FP

0.39

4.79

4.73

1.99

3.20

2.16

0.00

  

MRME (%)

65.52

60.72

62.90

56.98

60.66

56.64

17.89

 

5

TP

6.15

6.88

6.87

6.70

6.76

6.72

7.00

  

FP

0.38

5.28

5.27

1.82

3.04

2.14

0.00

  

MRME (%)

46.56

60.39

61.09

49.18

52.93

48.03

17.89

0.8

1

TP

2.07

4.12

4.05

3.30

3.76

3.34

7.00

  

FP

0.58

4.47

4.20

2.25

3.24

2.21

0.00

  

MRME (%)

58.85

44.01

45.63

44.16

42.13

42.63

18.04

 

3

TP

4.54

6.36

6.32

5.76

6.07

5.78

7.00

  

FP

0.52

5.16

4.88

2.03

2.90

2.37

0.00

  

MRME (%)

65.93

60.92

63.41

58.36

59.99

56.21

18.04

 

5

TP

5.90

6.85

6.83

6.51

6.61

6.59

7.00

  

FP

0.49

5.59

5.37

1.73

2.83

2.27

0.00

  

MRME (%)

52.57

59.45

62.29

53.52

56.85

50.16

18.04

  1. Oracle estimator is the least squares estimator of the true model, which contains seven nonzero covariates. TP (True Positive) is the average number of nonzero covariates being correctly selected. FP (False Positive) is the average number of zero covariates being incorrectly selected. Median relative model error (MRME) is used to measure the overall performance of different models