| Popis: |
Missing data is a common problem in clinical study. Removing missing data with complete case analysis (CCA) could lower power and bias the statistical conclusion. A variety of approaches have been used to deal with missing data. Several basic imputation methods would be introduced in this study. With the datasets derived from Teen-Longitudinal Assessment of Bariatric Surgery (Teen-LABS) study, group mean imputation (Gmean), total mean imputation (Tmean), expectation-maximization (EM) Algorithm, Markov chain Monte Carlo (MCMC) and fully conditional specification (FCS) imputation methods will be compared. Mean absolute error (MAE), root mean squared error (RMSE), mean of parameter bias, standard error of parameter bias will be used as evaluation criteria. Our results suggest that FCS is the rigorous statistical procedure for rare event data with high missing rate and binary outcome, which deserves more application in practice. |
