Skip to main content

Advertisement

Table 1 Scenarios investigated in the simulations (each investigated with 20, 40, 60 and 80% missing outcome data)

From: Multiple imputation using linked proxy outcome data resulted in important bias reduction and efficiency gains: a simulation study

    Factor 3: Change in Pr(IQ observed) for one SD increase in IQ Factor 2: Correlation between IQ and linked attainment score (KS4)
0.1 0.3 0.5 0.7 0.9
Main set of scenarios (each at 20, 40, 60, 80% missing IQ (factor 1)): 64 scenarios
    0      
    0.05
    0.1
    0.2
Secondary sets of scenarios (each at 20, 40, 60, 80% missing IQ (factor 1)): 36 scenarios
Missing linked data Factor 5: Change in Pr(KS4 observed) for one SD increase in KS4 Factor 4: Association between IQ and Pr(IQ observed) dependent on breastfeeding?       
No Yes 0.1a
Yes, 20% − 0.1 No 0.1     
  + 0.1   0.1     
  1. aIn the baseline breastfeeding group; reduction in this coefficient of 0.025 for each consecutive breastfeeding group