Term: Multiple Imputation
Last Updated: 2011-10-24
A technique used to overcome the problem of missing data, which increases statistical power for finding associations without artificially reducing the variation in the data. Multiple imputation does not simply insert a value for each missing data point or replace it with the mean of the other data points. Rather, it creates a set of imputed values for the missing data in a way that ensures that the variance/covariance structure present within the collected data remains the same. Multiple imputation can be used to overcome the problem of missing data for some questions in some surveys, for example, if a certain question was asked in most but not all survey waves.