Comparison of Procedures to Account for Certainty Primary Sampling Units

• Activity Number: 441
• Type: Contributed
• Date/Time: Thursday, August 7, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Bidisha Ghosh*+ and Frank J. Potter and Nuria Diaz-Tena and Don Sig Jang
• Companies: Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc.
• Keywords: variance estimation ; BRR ; Taylor series linearization ; certainty PSUs

Abstract

For the balanced repeated replication (BRR) variance estimator, various procedures have been used for primary sampling units (PSUs) selected with certainty. These include pairing PSUs units as pseudo-strata or splitting certainty PSUs to form two pseudo-PSUs. For the National Survey of SSI Children and Families (NSCF), we developed procedures for variance estimation using both BRR and Taylor series linearization. The NSCF has a complex multi-stage sample design with 75 PSUs selections with probability proportion to a composite size measure. Among the 75 PSU selections, 68 were non-certainty selections and 6 were certainty selections (one PSU had two selections). For BRR, noncertainty PSUs were paired to form 34 pseudo-strata. For the certainty PSUs, we explored alternative procedures to form pseudo-strata and pseudo-PSUs. For each procedure, we used Fay's method in computing the replicate weights, that is the weights of complementary PSUs were multiplied by k and 2-k for an appropriate choice of a scalar k. Our study compares the BRR sampling variances using different procedures to allocate the certainty PSUs to the sampling variance using the Taylor series linearization.