Skip to main content

Table 10 Comparison of approaches for the estimation of natural effects on the odds ratio scale for Scenario 5 with a binary mediator (based on 1000 data sets of size \(n=1000\))

From: An empirical evaluation of approximate and exact regression-based causal mediation approaches for a binary outcome and a continuous or a binary mediator for case-control study designs

Effect

Approach

True value

Mean

Bias

Relative bias (%)

SD

RMSE

CP (%) delta

CP (%) boot

NDE

Approx_Naive

3.023

3.949

0.926

30.648

0.596

1.101

61.8

58.2

Approx_C

3.023

3.808

0.785

25.979

0.596

0.986

71.7

69.1

Approx_IPW

3.023

3.870

0.847

28.034

0.599

1.038

67.8

63.9

Exact_Naive

3.023

2.529

-0.494

-16.346

0.560

0.747

83.3

85.6

Exact_IPW

3.023

3.063

0.040

1.336

0.533

0.535

95.0

94.6

Unified

3.023

3.811

0.788

26.058

0.598

0.989

72.8

-

NIE

Approx_Naive

2.213

1.838

-0.375

-16.948

0.134

0.398

23.5

21.5

Approx_C

2.213

1.785

-0.428

-19.327

0.130

0.447

23.0

16.9

Approx_IPW

2.213

1.842

-0.371

-16.752

0.135

0.395

29.7

25.6

Exact_Naive

2.213

2.759

0.546

24.670

0.518

0.752

77.5

78.5

Exact_IPW

2.213

2.238

0.025

1.148

0.266

0.268

96.6

94.9

Unified

2.213

1.780

-0.433

-19.566

0.129

0.452

15.0

-

TE

Approx_Naive

6.690

7.233

0.543

8.124

1.053

1.185

93.8

92.3

Approx_C

6.690

6.762

0.073

1.087

0.934

0.937

97.0

95.6

Approx_IPW

6.690

7.099

0.409

6.119

1.022

1.100

95.2

94.1

Exact_Naive

6.690

6.761

0.071

1.063

0.933

0.935

96.0

95.6

Exact_IPW

6.690

6.773

0.083

1.239

0.942

0.945

96.6

95.5

Unified

6.690

6.746

0.056

0.844

0.930

0.932

95.9

-

  1. boot Bootstrap, CP Coverage probability, NDE Natural direct effect, NIE Natural indirect effect, RMSE Root mean squared error, SD Standard deviation, TE Total effect