Skip to main content

Table 4 Performance of each method in terms of number of signals, False Discovery Proportion (FDP), specificity and sensitivity. Operating characteristics are calculated based on drugs with known status

From: New adaptive lasso approaches for variable selection in automated pharmacovigilance signal detection

Method

Number of

Number of signals

Number of false

FDP

Specificity

Sensitivity

 

generated signals

with known status

positive signals

(%)

(%)

(%)

  

(positive or negative)

    

RFET

249

82

13

15.9

93.6

51.9

lasso-cv

220

79

5

6.3

97.5

55.6

lasso-bic

170

65

5

7.7

97.5

45.1

lasso-perm

158

60

4

6.7

98.0

42.1

CISL

109

48

2

4.2

99.0

34.6

adapt-cv

188

70

5

7.1

97.5

48.9

adapt-univ

179

65

4

6.2

98.0

45.9

adapt-univ-bic

163

61

4

6.6

98.0

42.9

adapt-bic

151

60

4

6.7

98.0

42.1

adapt-cisl

153

60

2

3.3

99.0

43.6

ps-adjust

209

71

6

8.5

97.0

48.9

ps-mw

86

40

0

0.0

100.0

30.1

ps-iptw

49

15

3

20.0

98.5

9.0

ps-iptwT

260

84

12

14.3

94.1

54.1