Skip to main content

Table 3 Diagnostic measures for imputation methods

From: Dealing with missing data in a multi-question depression scale: a comparison of imputation methods

Missing Data Scenario

Method

Mean

SD

Spearman

% Misclassified

Kappa

P = 0.10

N = 1379**

μ = 43.68

σ = 10.98

Random Selection

45.99*

10.65

0.906

15% (207)

0.684

 

Preceding Question

44.69*

10.07

0.946

8.7% (120)

0.807

 

Question Mean

43.75

9.84*

0.986

7.5% (104)

0.823

 

Individual Mean

43.74

11.11

0.986

5.4% (74)

0.880

 

Single Regression

44.03

10.71

0.981

5.6%(77)

0.873

 

Multiple Imputation

44.01

10.73

0.987

4.7% (65)

0.893

P = 0.20

N = 1562**

μ = 43.64

σ = 10.98

Random Selection

47.25*

11.14

0.784

28.2% (440)

0.452

 

Preceding Question

46.41*

9.79*

0.898

14.4% (225)

0.700

 

Question Mean

43.59

8.88*

0.974

12.1% (189)

0.709

 

Individual Mean

43.59

11.26

0.974

8.9% (139)

0.802

 

Single Regression

44.03

10.65

0.966

9.6% (150)

0.781

 

Multiple Imputation

44.06

10.49

0.976

7.0% (110)

0.839

P = 0.30

N = 1579**

μ = 43.62

σ = 10.93

Random Selection

49.09*

11.92*

0.610

41.0% (647)

0.267

 

Preceding Question

48.62*

9.55*

0.867

23.6% (373)

0.549

 

Question Mean

43.60

8.05*

0.958

14.9% (235)

0.629

 

Individual Mean

43.66

11.33

0.955

10.8% (171)

0.760

 

Single Regression

44.39

10.33

0.937

11.4%(180)

0.738

 

Multiple Imputation

44.32

10.21

0.959

9.2% (145)

0.789

Q6

N = 1406**

μ = 43.49

σ = 10.89

Random Selection

45.62

10.38

0.901

16.6% (233)

0.649

 

Preceding Question

41.66*

10.73

0.970

10.2% (143)

0.753

 

Question Mean

43.43

9.67*

0.987

8.4% (118)

0.798

 

Individual Mean

43.37

11.03

0.984

5.7% (80)

0.870

 

Single Regression

43.66

10.67

0.978

6.8%(95)

0.842

 

Multiple Imputation

43.67

10.61

0.986

5.8% (81)

0.866

MAR – Age and Sex

N = 1429**

μ = 43.60

σ = 10.90

Random Selection

45.85

10.48

0.885

18.1 %(259)

0.618

 

Preceding Question

44.81

10.09

0.940

8.9% (127)

0.804

 

Question Mean

43.63

9.65*

0.984

7.4% (106)

0.825

 

Individual Mean

43.65

11.05

0.982

5.7% (82)

0.867

 

Single Regression

43.89

10.67

0.978

7.1%(102)

0.835

 

Multiple Imputation

43.91

10.58

0.985

5.3% (77)

0.877

MNAR

N = 1406**

μ = 43.51

σ = 10.80

Random Selection

45.82*

10.46

0.899

15.7% (221)

0.741

 

Preceding Question

44.44

10.06

0.947

9.7% (136)

0.839

 

Question Mean

43.51

9.66*

0.987

8.4% (118)

0.850

 

Individual Mean

43.50

10.90

0.985

5.9% (83)

0.902

 

Single Regression

43.54

10.78

0.975

7.7% (108)

0.871

 

Multiple Imputation

43.54

10.65

0.986

6.1% (86)

0.897

  1. * significant difference from the population statistics at 95% confidence
  2. ** Participants for which no observations were randomly deleted are excluded from the analysis. When there are no missing values to impute, the calculated score is the same as the known "true" score thus the scores correlate perfectly (spearman = 1.0)