Skip to main content

Table 2 Mean errors between observed and imputed BMI values, three sequential missing values per individual (interpolation only)

From: Longitudinal multiple imputation approaches for body mass index or other variables with very low individual-level variability: the mibmi command in Stata

Cases

Methoda

Obs.

Mean

Std.Dev

Min

Max

Allb

Simple ×1

30,000

0.980

1.002

0.000

16.017

mibmi ×100

30,000

0.989

1.004

0.000

16.034

Twofold ×100

30,000

1.137

1.155

0.000

18.318

Time point 1

Simple ×1

10,000

0.935

0.945

0.000

9.829

mibmi ×100

10,000

0.943

0.947

0.000

9.779

Twofold ×100

10,000

1.094

1.114

0.000

18.318

Time point 2

Simple ×1

10,000

1.059

1.068

0.000

16.017

mibmi ×100

10,000

1.069

1.071

0.000

16.034

Twofold ×100

10,000

1.234

1.231

0.000

17.126

Time point 3

Simple ×1

10,000

0.947

0.984

0.000

10.645

mibmi ×100

10,000

0.955

0.985

0.000

10.538

Twofold ×100

10,000

1.084

1.111

0.000

13.473

  1. aSimple refers to a single imputation that ignores variability in the observations (option xsimp); mibmi refers to the default multiple imputation approach with the command and 100 imputations; twofold refers to the twofold algorithm described in the paper and 100 imputations
  2. bAll refers to aggregates across all three time points