Skip to main content

Advertisement

Table 2 Estimates of associations between predictors and health outcome at 50% response-rate

From: Preventing bias from selective non-response in population-based survey studies: findings from a Monte Carlo simulation study

bnon   10% most extreme values on the health outcome totally missing Extreme values on the health outcome not totally missing
CC FIML MI CC FIML MI
×1,×2,× 3 Y Pred. Bpred (SE) 95% coverage Bpred (SE) 95% coverage Bpred (SE) 95% coverage Bpred (SE) 95% coverage Bpred (SE) 95% coverage Bpred (SE) 95% coverage
0.3 0.0 ×1 .18 (.05) 91 .18 (.05) 92 .17 (.05) 88 .20 (.04) 92 .20 (.05) 92 .20 (.05) 93
   ×2 .25 (.05) 91 .26 (.05) 93 .25 (.05) 76 .29 (.04) 94 .30 (.04) 94 .29 (.05) 93
0.1 0.1 ×1 .17 (.05) 89 .17 (.05) 91 .17 (.05) 88 .20 (.04) 92 .20 (.04) 92 .19 (.05) 93
   ×2 .25 (.04) 89 .26 (.05) 93 .25 (.05) 82 .29 (.04) 96 .29 (.04) 95 .29 (.05) 95
0.3 0.1 ×1 .16 (.05) 88 .17 (.05) 90 .16 (.05) 84 .18 (.05) 92 .18 (.05) 92 .18 (.05) 90
   ×2 .24 (.05) 84 .25 (.05) 89 .23 (.05) 69 .27 (.04) 92 .28 (.05) 95 .27 (.05) 86
0.0 0.3 ×1 .18 (.05) 90 .18 (.05) 93 .17 (.05) 91 .20 (.04) 94 .20 (.04) 94 .19 (.05) 95
   ×2 .26 (.04) 90 .26 (.05) 94 .26 (.05) 85 .29 (.04) 95 .29 (.04) 93 .29 (.05) 95
0.1 0.3 ×1 .16 (.05) 88 .17 (.05) 91 .16 (.05) 86 .18 (.05) 92 .18 (.05) 93 .18 (.05) 92
   ×2 .24 (.04) 82 .25 (.05) 91 .24 (.05) 78 .27 (.04) 93 .28 (.04) 95 .27 (.05) 92
0.3 0.3 ×1 .12 (.05) 68 .12 (.05) 69 .11 (.05) 54 .13 (.05) 70 .13 (.05) 72 .12 (.05) 60
   ×2 .20 (.05) 58 .21 (.05) 68 .19 (.05) 43 .21 (.05) 68 .22 (.05) 76 .20 (.05) 49
  1. The true population values are bpred ×1 = 0.20, bpred ×2 = 0.30. Different degrees of dependency between study variables and non-response are modeled
  2. Y = health outcome. bnon = regression coefficients of normally distributed liability of non-response (L) on predictors (× 1, × 2, x3) and on health outcome
  3. bpred coefficients for the regression of health outcome on × 1 and × 2, when the outcome is treated as a categorical variable, and the probit-link is used
  4. SE standard error, 95% coverage percentage of the randomly drawn samples providing a 95% confidence interval containing the true population value. FIML full information maximum likelihood, MI multiple imputation (predictive mean matching). N in the original sample before non-response = 1000. X3 was included as auxiliary variable in FIML and as predictor in MI. 50 data sets were imputed