Skip to main content

Table 3 Empirical coverage of 95% confidence intervals for β2 after multiple imputation by PMM

From: Multiple imputation by predictive mean matching in cluster-randomized trials

ICC

Cluster size

Clusters

PMM

PMM

PMM

PMM

PMM

PMM

ρ

m

k

FE

dist

RE

draw

avg

IGN

0.03

4

100

91.5

94.4

94.0

95.1

94.5

94.7

 

8

10

94.5

94.1

94.9

95.1

93.5

94.7

  

20

92.6

93.8

93.7

93.9

93.2

94.2

  

40

93.1

93.8

94.7

94.8

94.0

95.1

 

40

10

97.3

96.1

97.0

97.0

94.3

97.3

  

20

95.8

93.7

94.6

95.1

91.7

95.1

  

40

95.4

94.8

94.8

94.6

92.1

95.3

 

400

4

100

100

100

100

99.7

100

0.08

4

100

91.8

94.2

93.2

95.4

93.8

94.8

 

8

10

94.0

94.3

94.5

95.0

92.5

94.8

  

20

92.4

93.3

94.3

94.5

91.9

94.0

  

40

93.3

92.7

93.8

95

92.2

95.0

 

40

10

97.3

96.5

97

96.8

93.0

97.3

  

20

95.5

94.5

95.5

94.8

90.4

95.2

  

40

95.2

95.1

94.9

94.8

91.2

95.6

 

400

4

100

100

100

100

100

100

Bolded values indicate coverage rates below 95%.

Italicized values indicate coverage rates above 95%.

Missing data were generated completely at random (MCAR) with a response rate of 60%, and data were imputed using an imputation model that correctly specified the covariate effects.