In this report we consider variance estimation for tables in the Dutch virtual Census, when mass imputation is used to predict all missing values of educational attainment in the population.
Two alternative approaches are developed to estimate the variance of frequency tables in the Dutch virtual Census, when all missing values of educational attainment in the population are predicted by mass imputation. The first approach is based on an analytical variance approximation; the second approach is a bootstrap method. Both approaches are tested in a small simulation study.
