Skip to main content

Table 4 Relative bias percentage, RMSE and coverage probability for cannabis 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%

52.0%

0.00046

0.09%

0.0299

95.6%

 LR

10%

52.0%

0.00042

0.08%

0.0285

95.8%

 HD

10%

52.0%

0.00033

0.06%

0.0253

94.2%

 MI (M = 5)

10%

52.0%

0.00056

0.11%

0.0328

95.6%

 MI (M = 20)

10%

52.0%

0.00050

0.10%

0.0311

96.2%

 LD

30%

52.0%

0.00021

0.04%

0.0201

94.8%

 LR

30%

52.0%

0.00029

0.06%

0.0238

93.0%

 HD

30%

52.0%

0.00034

0.07%

0.0256

92.0%

 MI (M = 5)

30%

52.0%

0.00032

0.06%

0.0247

95.8%

 MI (M = 20)

30%

52.0%

0.00036

0.07%

0.0262

95.6%

 LD

50%

52.0%

0.00070

0.14%

0.0368

94.2%

 LR

50%

52.0%

0.00072

0.14%

0.0373

91.0%

 HD

50%

52.0%

0.00057

0.11%

0.0332

87.4%

 MI (M = 5)

50%

52.0%

0.00064

0.12%

0.0351

95.4%

 MI (M = 20)

50%

52.0%

0.00079

0.15%

0.0389

95.0%

Missing Data Mechanism: MAR

 LD

10%

51.0%

-0.00991

-1.91%

0.0145

85.4%

 LR

10%

52.0%

0.00032

0.06%

0.0102

95.8%

 HD

10%

51.8%

-0.00134

-0.26%

0.0107

93.6%

 MI (M = 5)

10%

52.0%

0.00027

0.05%

0.0101

96.2%

 MI (M = 20)

10%

52.0%

0.00034

0.07%

0.0102

96.2%

 LD

30%

49.5%

-0.02443

-4.70%

0.0273

48.8%

 LR

30%

52.2%

0.00207

0.40%

0.0112

93.2%

 HD

30%

51.7%

-0.00241

-0.46%

0.0118

91.2%

 MI (M = 5)

30%

52.0%

0.00054

0.10%

0.0110

95.8%

 MI (M = 20)

30%

52.0%

0.00038

0.07%

0.0111

95.0%

 LD

50%

48.0%

-0.03969

-7.64%

0.0421

21.0%

 LR

50%

52.3%

0.00340

0.65%

0.0124

90.4%

 HD

50%

51.7%

-0.00280

-0.54%

0.0137

86.0%

 MI (M = 5)

50%

52.0%

0.00029

0.06%

0.0122

95.0%

 MI (M = 20)

50%

52.0%

0.00028

0.05%

0.0122

94.4%

Missing Data Mechanism: MNAR

 LD

10%

49.9%

-0.02029

-3.90%

0.0230

53.8%

 LR

10%

50.4%

-0.01557

-3.00%

0.0188

69.0%

 HD

10%

50.5%

-0.01480

-2.85%

0.0183

70.4%

 MI (M = 5)

10%

50.5%

-0.01495

-2.88%

0.0183

72.6%

 MI (M = 20)

10%

50.5%

0.47888

92.16%

0.0183

72.2%

 LD

30%

46.1%

-0.05821

-11.20%

0.0596

0.0%

 LR

30%

47.7%

-0.04286

-8.25%

0.0444

1.8%

 HD

30%

47.9%

-0.04112

-7.91%

0.0429

3.6%

 MI (M = 5)

30%

47.9%

-0.04078

-7.85%

0.0423

4.0%

 MI (M = 20)

30%

-1.5%

-0.04075

-7.84%

0.0423

4.0%

 LD

50%

42.2%

-0.09781

-18.82%

0.0988

0.0%

 LR

50%

45.9%

-0.06050

-11.64%

0.0616

0.0%

 HD

50%

46.1%

-0.05828

-11.22%

0.0597

0.0%

 MI (M = 5)

50%

46.3%

-0.05702

-10.97%

0.0582

0.0%

 MI (M = 20)

50%

0.0%

0.00013

0.02%

0.0582

0.0%

  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