Replication Variance Estimation for Imputed Data

Replication methods are used frequently to con-struct variance estimators under complex sampling designs. We develop a replication variance estima-tor for the situation in which some of the sample elements are unobserved and their information is im-puted. The method relies on fractional hot-deck im-putation and adjusts the variance replicates to ac-count for the imputation mechanism. The properties of the method are discussed