Skip to main content

Table 6 MNAR and efficacy rate 98 % versus 95 % (RD 0.030): estimated efficacy differences, coverage and bias for 5 %, 15 % and 30 % averages of number of simulated data sets that converged of the 5000 data sets, 50 imputations

From: Is using multiple imputation better than complete case analysis for estimating a prevalence (risk) difference in randomized controlled trials when binary outcome observations are missing?

Model No. of data sets* RD (RMSE) Coverage Bias
All outcomes recorded: 5,000 0.030 (0.026) 0.939 0.000
5 % of outcomes missing     
CC 5,000 0.031 (0.027) 0.947 +0.001
MI: wt, hb, age, para 4,989 0.031 (0.028) 0.960 +0.001
MI: hb, age, para 4,988 0.030 (0.027) 0.961 0.000
MI: hb, age, para, group 4,992 0.032 (0.028) 0.950 +0.002
MI: wt, hb, age, para, group 4,992 0.032 (0.028) 0.955 +0.002
15 % of outcomes missing     
Complete Case 5,000 0.035 (0.030) 0.951 +0.005
MI: wt, hb, age, para 4,985 0.029 (0.031) 0.981 0.000
MI: hb, age, para 4,990 0.030 (0.031) 0.981 -0.001
MI: hb, age, para, group 4,991 0.035 (0.032) 0.960 +0.005
MI: wt, hb, age, para, group 4,989 0.036 (0.032) 0.961 +0.006
30 % of outcomes missing     
CC 5,000 0.042 (0.036) 0.950 +0.012
MI: wt, hb, age, para 4,982 0.030 (0.038) 0.992 0.000
MI: hb, age, para 4,991 0.030 (0.037) 0.991 0.000
MI: hb, age, para, group 4,988 0.043 (0.040) 0.965 +0.013
MI: wt, hb, age, para, group 4,991 0.043 (0.041) 0.965 +0.013
  1. *Number of data sets for which convergent analysis was achieved