Rubin’s approach, called multiple imputation, takes that distribution into account. To use it, first make several copies of your data set. For a given missing value in one copy, randomly assign a ...