Skip to main content

Table 2 Relative bias, RMSE and coverage probability for heroin use comparing validation data to imputed data

From: Comparing single and multiple imputation strategies for harmonizing substance use data across HIV-related cohort studies

Method

% Missing

Estimate

Mean Bias

% Relative Bias

RMSE

Coverage

Missing Data Mechanism: MCAR

 LD

10%

3.0%

0.00003

0.11%

0.0036

94.6%

 LR

10%

3.0%

0.00006

0.19%

0.0036

94.6%

 HD

10%

3.0%

0.00001

0.04%

0.0036

94.0%

 MI (M = 5)

10%

3.0%

0.00034

1.14%

0.0036

94.8%

 MI (M = 20)

10%

3.0%

0.00033

1.09%

0.0036

95.0%

 LD

30%

3.0%

0.00009

0.29%

0.0042

94.6%

 LR

30%

3.0%

0.00010

0.34%

0.0038

92.2%

 HD

30%

3.0%

0.00006

0.19%

0.0042

89.6%

 MI (M = 5)

30%

3.1%

0.00108

3.60%

0.0040

94.2%

 MI (M = 20)

30%

3.1%

0.00105

3.50%

0.0039

94.2%

 LD

50%

3.0%

-0.00007

-0.23%

0.0049

94.2%

 LR

50%

3.0%

-0.00006

-0.20%

0.0039

90.0%

 HD

50%

3.0%

0.00006

0.20%

0.0044

88.6%

 MI (M = 5)

50%

3.2%

0.00183

6.09%

0.0045

93.8%

 MI (M = 20)

50%

3.2%

0.00181

6.03%

0.0043

93.6%

Missing Data Mechanism: MAR

 LD

10%

2.0%

-0.00999

-33.33%

0.0105

14.0%

 LR

10%

2.9%

-0.00110

-3.66%

0.0037

91.8%

 HD

10%

2.9%

-0.00049

-1.65%

0.0037

92.8%

 MI (M = 5)

10%

3.0%

0.00031

1.03%

0.0037

95.0%

 MI (M = 20)

10%

3.0%

0.00034

1.12%

0.0037

94.2%

 LD

30%

1.4%

-0.01554

-51.84%

0.0158

1.0%

 LR

30%

2.7%

-0.00256

-8.55%

0.0044

83.4%

 HD

30%

2.9%

-0.00132

-4.40%

0.0043

87.2%

 MI (M = 5)

30%

3.1%

0.00118

3.95%

0.0042

95.4%

 MI (M = 20)

30%

3.1%

0.00121

4.02%

0.0041

95.2%

 LD

50%

1.1%

-0.01871

-62.41%

0.0189

0.0%

 LR

50%

2.7%

-0.00337

-11.25%

0.0051

75.6%

 HD

50%

2.8%

-0.00151

-5.03%

0.0048

82.8%

 MI (M = 5)

50%

3.2%

0.00201

6.69%

0.0048

93.6%

 MI (M = 20)

50%

3.2%

0.00205

6.83%

0.0046

93.8%

Missing Data Mechanism: MNAR

 LD

10%

1.6%

-0.01361

-45.41%

0.0139

0.8%

 LR

10%

1.8%

-0.01181

-39.41%

0.0122

3.4%

 HD

10%

1.9%

-0.01131

-37.73%

0.0117

7.0%

 MI (M = 5)

10%

1.9%

-0.01101

-36.74%

0.0114

11.4%

 MI (M = 20)

10%

1.9%

-0.01099

-36.67%

0.0114

10.8%

 LD

30%

1.2%

-0.01799

-60.03%

0.0182

0.0%

 LR

30%

1.8%

-0.01192

-39.77%

0.0123

3.2%

 HD

30%

1.9%

-0.01127

-37.59%

0.0117

6.8%

 MI (M = 5)

30%

2.0%

-0.01016

-33.91%

0.0106

20.6%

 MI (M = 20)

30%

2.0%

-0.01023

-34.13%

0.0106

18.6%

 LD

50%

1.0%

-0.02032

-67.82%

0.0205

0.0%

 LR

50%

2.0%

-0.00994

-33.15%

0.0105

13.8%

 HD

50%

2.1%

-0.00934

-31.16%

0.0101

22.8%

 MI (M = 5)

50%

2.3%

-0.00687

-22.91%

0.0077

69.8%

 MI (M = 20)

50%

2.3%

-0.00688

-22.97%

0.0077

65.8%

  1. Abbreviations, LD Listwise deletion, LR Logistic regression, HD Hot-deck, MI Multiple imputation, MCAR Missing completely at random, MAR Missing at random, MNAR Missing not at random, RMSE Root mean square error