ABSTRACT
Many modern surveys use computer-intensive methods to calculate variances of estimates. The survey analysis procedures in SAS® software will construct replicate weights for a survey with stratification and clustering, and will analyze survey data for which replicate weights are supplied. This chapter shows how to use random group methods to analyze survey data, and how to create replicate weights for the balanced repeated replication, jackknife, and bootstrap methods. Confidence intervals for quantiles may be produced either with replicate weight variance estimation or with the Woodruff method that is based on the confidence interval for an estimated proportion. Each concept is illustrated with code and output produced for examples in Chapter 9 of Sampling: Design and Analysis, Third Edition.
