A comparison of multiple imputation methods for handling missing values in longitudinal data in the presence of a time-varying covariate with a non-linear association with time: a simulation study

Table 3 Performance of various methods for handling 50% missingness in BMI for age z-scores; true OR^a = 1.1(log(OR) = 0.1)

Empirical SE empirical standard error, FCS fully conditional specification, MAR missing at random, MCAR missing completely at random, Model-based SE model based standard error, MVNI multivariate normal imputation, RMSE root mean square error, two-fold FCS two-fold fully conditional specification algorithm
^aTrue OR represents the true odds ratio between sleep problems and BMI for age z-scores
^bMonte Carlo standard error did not exceed 0.0006
^cResults for the two-fold FCS with a time window width of 1, that is, including immediately adjacent time points
^dResults for the two-fold FCS with a time window width of 2, that is, including two adjacent time points

ISSN: 1471-2288