This is a sophisticated method to estimate the impact of missing values in data, particularly in multivariate analyses of data. It can provide better reduction in bias than less sophisticated ways of handling data with missing values (which we will almost always have in any real world therapy data), including reporting analyses simply of the complete data. However, that superiority is only guaranteed when data is missing completely at random (MCAR) and where data are not MCAR it is at least theoretically possible for MICE based analyses to be as biased as cruder methods. Much more work is needed for us to be clear when and how MICE should be used for real world therapy change/outcome data.
Try also … #
Bias
Missing values
