Skip to main content

Table 1 Summary of empirical estimates of log-OR and relative bias (%), power (%), coverage (%), model standard error and empirical standard error derived from χ2 method and GLMM test with compound symmetric variance-covariance matrix from 1000 simulations with a total sample size N = 398

From: Power difference in a χ2 test vs generalized linear mixed model in the presence of missing data – a simulation study

 

pi1 = 0.1, ρ = 0.3

Log-odds (% bias)

Power (%)c

Coverage (%)c

Model SEc

Empirical SEc

Complete

χ2a

0.831 (0)

82.0

95.0

0.300

0.300

 GLMMb

0.886 (0)

85.7

96.8

0.301

0.283

MCAR

Equal drop out (30% drop out in each arm)

χ2

0.840 (1.1)

65.6

95.0

0.362

0.371

  GLMM

0.894 (1.0)

73.2

96.9

0.355

0.329

Unequal dropouts 1 (20% drop out in control, 40% drop out in treatment)

χ2

0.814 (−2.0)

65.8

95.4

0.356

0.356

  GLMM

0.886 (0.1)

72.8

96.4

0.353

0.326

Unequal dropouts 2 (40% drop out in control, 20% drop out in treatment)

χ2

0.841 (1.3)

62.5

95.5

0.375

0.380

  GLMM

0.886 (0)

71.5

97.1

0.362

0.326

MAR1 (observations more likely to be deleted if prior observation was positive)

 Equal drop out (30% drop out in each arm)

χ2

0.921 (10.9)

69.2

95.2

0.386

0.395

  GLMM

0.949 (7.2)

76.5

96.7

0.363

0.337

Unequal dropouts 1 (20% drop out in control, 40% drop out in treatment)

χ2

0.834 (0.5)

62.8

94.9

0.376

0.386

  GLMM

0.928 (4.8)

74.4

96.8

0.359

0.331

Unequal dropouts 2 (40% drop out in control, 20% drop out in treatment)

χ2

0.991 (19.3)

74.4

95.7

0.402

0.407

  GLMM

0.959 (8.3)

76.7

97.4

0.369

0.340

MAR2 (observations less likely to be deleted if prior observation was positive)

Equal drop out (30% drop out in each arm)

χ2

0.791 (−4.8)

65.5

95.0

0.353

0.358

  GLMM

0.858 (−3.1)

71.1

97.1

0.354

0.327

Unequal dropouts 1 (20% drop out in control, 40% drop out in treatment)

χ2

0.848 (2.1)

70.9

94.7

0.351

0.362

  GLMM

0.871 (−1.6)

71.9

96.7

0.353

0.328

Unequal dropouts 2 (40% drop out in control, 20% drop out in treatment)

χ2

0.749 (−9.9)

56.2

93.0

0.361

0.369

  GLMM

0.853 (−3.7)

67.8

95.7

0.359

0.337

  1. aIndependent two-sample χ2 test for test of treatment effect at time point 3
  2. bGeneralized linear mixed model with compound symmetric variance-covariance matrix, contrast used to estimate treatment effect at time point 3
  3. cRange of Monte Carlo SE: Power (0.011–0.016); Coverage (0.005–0.008); Model SE (0.0004–0.002); Empirical SE (0.006–0.009)
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us