The process of generated several different data imputations and then pooling then so that the standard error includes a contribution from the difference between imputations.