Replicate methods for variance estimation are commonly used in large sample surveys with many variables. The procedure uses estimators computed on subsets of the sample, where subsets are selected in a way that reflects the sampling variability. Replication variance estimation is an appealing alternative to Taylor linearization variance estimation for nonlinear functions. Replicate methods have the advantage of transferring the complexity of variance estimation from data set end users to the statistician working on creating the output data set. By providing weights for each subset of the sample, called replication weights, end users can estimate the variance of a large variety of nonlinear estimators using standard weighted sums. Jackknife, balanced half-samples, and bootstrap methods are three main replication variance methods used in sample surveys. The ...
Looks like you do not have access to this content.