Table 2. Simulation study: methods for handling missing values in partially observed components z2 and z3. y, composite endpoint; x, randomized treatment; z1, fully observed component.
| Method | Variable(s) imputed | Imputation model predictors |
|---|---|---|
| CRA | ||
| Deriv | ||
| aMI-CRA | y CRA | x |
| aMI-Deriv | y deriv | x |
| bMIC-main | z2, z3 | z1, z2 or z3, x |
| bMIC-x | z2, z3 | z1, z2 or z3; stratified by x |
| bMIC-x-z1 | z2, z3 | z2 or z3; stratified by z1 and x |
a
Univariate MI using logistic regression.
b
MICE using logistic regression for conditional models.