Skip to main content

Table 2 Comparison of the absolute value of the bias (|Bias|), the variance and the MSE of estimating the PAF among the logistic regression approach (logit), the conventional B-splines approach (conB), and the developed approach (monB)

From: Semiparametric estimation of the attributable fraction when there are interactions under monotonicity constraints

  

|Bias|(×10−1)

Variance (×10−2)

MSE (×10−2)

Sample size

Approach

A

B

C

A

B

C

A

B

C

n=100

logit

1.63

3.62

8.75

10.80

37.19

94.47

13.45

50.24

170.94

 

logit ∗

1.86

1.28

2.40

7.98

12.54

9.53

11.44

14.17

15.29

 

conB ∗

1.74

2.84

2.64

5.18

10.09

12.86

8.20

18.17

19.80

 

monB

1.48

1.53

1.65

3.69

4.94

5.25

5.88

7.28

7.96

n=200

logit

2.07

3.67

8.53

6.27

26.80

68.97

10.53

40.25

141.63

 

logit ∗

2.14

1.74

2.85

5.35

9.79

6.96

9.93

12.81

15.09

 

conB ∗

0.33

1.43

2.78

5.43

7.95

11.02

5.54

9.99

18.74

 

monB

1.16

1.00

1.17

2.75

3.20

3.88

4.08

4.20

5.26

  1. The logit ∗ and conB ∗ estimates are obtained by censoring the original estimates at 0 or at 1
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us