Multiple imputation for handling missing outcome data when estimating the relative risk

Table 3 Bias in Scenarios with X ₁ and X ₂ Continuous, Coordinated Missing Data Mechanism and Outcome Prevalence = 0.30

Simulation scenario	Parameter	MVNI	MVNI + deletion	FCS	FCS + deletion
1. Corr(X ₁, X ₂) = 0.70, β ₁ = β ₂ = log(3), λ = 2	β ₁	−0.26	0.01	−0.09	0.02
1. Corr(X ₁, X ₂) = 0.70, β ₁ = β ₂ = log(3), λ = 2	β ₂	−0.27	−0.11	−0.24	−0.12
2. As in (1.), but with Corr(X ₁, X ₂) = 0	β ₁	−0.27	−0.05	−0.15	−0.05
2. As in (1.), but with Corr(X ₁, X ₂) = 0	β ₂	−0.21	−0.06	−0.16	−0.06
3. As in (1.), but with β ₁ = 0	β ₁	−0.04	0.02	0.01	0.02
	β ₂	−0.17	−0.03	−0.10	−0.03
4. As in (1.), but with β ₂ = 0	β ₁	−0.24	0.00	−0.10	0.00
	β ₂	0.00	0.00	0.00	0.00
5. As in (1.), but with β ₁ = β ₂ = 0	β ₁	−0.01	−0.01	−0.01	−0.01
	β ₂	0.01	0.01	0.01	0.01
6. As in (1.), but with λ = 0 (MCAR)	β ₁	−0.11	0.00	0.00	0.00
	β ₂	−0.17	−0.08	−0.08	−0.08

Abbreviations: Corr correlation, MCAR missing completely at random, MVNI multivariate normal imputation, FCS fully conditional specification

ISSN: 1471-2288