Table 2. Advantages and disadvantages of methods to tackle missing data.

Feature	Complete-case	Median substitution	EM	MICE
No special software is needed	Yes	Yes	Yes	No
Easy to communicate with clinical audience	Yes	Yes	Yes	No
Do not require distributional assumption	Yes	Yes	No	Yes
Preserve data characteristics	No	No	Yes	Yes
Convergence of imputation model is not an issue	Yes	Yes	No	No
Takes imputation uncertainty into account	No	No	No	Yes
Any particular problem	Diminishes the power Gives biased estimated if not MCAR	Artificially reduces the variance	Might give out of range estimates	Requires aggregation of estimates