ABSTRACT
Missing data are common in scientific research and daily life. In clinical trials, missing data can be caused by many different variables, including but not limited to patient refusal to continue in the study, treatment failures, data entry errors, adverse events, or patient relocations. In many medical settings missing data can cause difficulties in estimation, precision and inference. A completer analysis can reduce the power of hypothesis testing due to a reduction in sample size; on the other hand, non-completers might be more likely to have extreme values. Monotonic missingness is a common and simple pattern. Multiple imputation (MI) methods generate multiple copies of the “complete data set” by replacing missing values with randomly generated values using Bayesian or other methods, and analyze them as complete sets. The MI procedure creates missing-imputed data sets and stores them in the Outmi dataset.
