Skip to main content

Table 2 Comparison of the performance of different estimation methods of the effect of cumulative exposure to psychosocial stressors at work on blood pressure in the nonparametric plasmode simulation scenario

From: Longitudinal plasmode algorithms to evaluate statistical methods in realistic scenarios: an illustration applied to occupational epidemiology

 

Intermittent

Cessation

Onset

Chronic

Bias

    IPTW

-0.51

0.04

-0.38

-0.34

    g-computation

-0.59

-0.42

-0.39

-0.14

    TMLE

-0.20

-0.75

-0.02

-0.91

    TMLE+SL

-0.24

-0.73

-0.05

-0.90

    Baseline adjustment

-0.94

-0.77

0.58

-0.90

    Baseline and Time-varying covariates adjustment

-0.88

-0.82

0.04

-1.08

Standard deviation

    IPTW

1.02

1.49

1.16

2.93

    g-computation

0.49

0.80

0.66

0.98

    TMLE

0.90

0.84

0.63

1.43

    TMLE+SL

0.90

0.84

0.63

1.42

    Baseline adjustment

0.65

0.89

0.64

1.45

    aseline and Time-varying covariates adjustment

0.65

0.85

0.70

1.50

RMSE

    IPTW

1.14

1.49

1.22

2.95

    g-computation

0.77

0.90

0.77

0.99

    TMLE

0.92

1.12

0.63

1.70

    TMLE+SL

0.93

1.12

0.64

1.68

    Baseline adjustment

1.14

1.15

0.90

1.75

    Baseline and Time-varying covariates adjustment

1.09

1.21

0.65

1.81

Coverage of the 95% confidence interval

    IPTW

94%

91%

93%

71%

    g-computation

73%

89%

89%

92%

    TMLE

93%

83%

94%

86%

    TMLE+SL

93%

83%

94%

87%

    Baseline adjustment

70%

74%

80%

67%

    Baseline and Time-varying covariates adjustment

71%

69%

93%

65%

  1. Legend: IPTW = Inverse probability of treatment weighting, RMSE = root mean squarred error, SL = Super Learner, TMLE = Targeted maximum likelihood