252 – Replicate Weights and Variance Estimation
Jackknife and Other Replication Methods with a Reduced Number of Replicates
Stephen Ash
U.S. Census Bureau
Replicate weights used with surveys are becoming more available and expected for public use data. Data users can use replicate weights to estimate their own variances for point estimates or use them with sophisticated software for multivariate analyses. This paper will discuss how to produce replicate weights with a reduced number replicates for the jackknife and other replication estimators including using balanced repeated replication and successive difference replication. Our focus will be the use of these estimators with systematic random sampling from an ordered list. The reduced number of replicates is needed because data users require a reasonable number of replicates in their analysis. The paper will compare all of the replication estimators and examine how the reduction of replicates impacts the resultant variance estimators.