|
Abstract:
|
The widespread use of multiple imputation (MI) to address missing data has improved the quality of clinical trial reporting, but has also introduced the nuisance of estimates that are subject to Monte Carlo variation. It can be somewhat of a balancing act to ensure that enough imputations are used to guarantee precise results for decision making, while at the same time keeping the number of imputations low enough to avoid computational bottlenecks in the critical period after treatment unblinding. One way to approach this trade-off in a systematic manner is to require estimates to be stable, in the sense that another round of MI would yield the same results with high probability, to the chosen reporting precision. We show how this stability notion can be used to establish rules of thumb for making an informed choice of the number of imputations at an early stage of a clinical trial.
|
