The 131st Annual Meeting (November 15-19, 2003) of APHA

Abstract #63170

Frank J. Potter, PhD¹, Nuria Diaz-Tena, PhD¹, Stephen Williams, MS¹, Elizabeth Schaefer², and James Reschovsky, PhD². (1) Mathematica Policy Research, Inc, P.O. Box 2393, Princeton, NJ 08543, 609-936-2799, FPOTTER@Mathematica-MPR.COM, (2) Center for Studying Health System Change, 600 Maryland Avenue, SW, Washington, DC 20024

The Community Tracking Study (CTS) includes national surveys of persons and physicians using a complex sampling design. The CTS surveys are designed primarily for tracking changes in the health system at a community level (health care market areas) and also support national estimates for key subpopulations. For national estimates, the health care market areas are used as the primary sampling units (PSUs). Since nearly 25 percent of these PSUs are selected without replacement and with unequal probabilities using Chromy’s sequential selection algorithm, we computed joint inclusion probabilities and use the Yates-Grundy-Sen variance estimator for the national estimates. Because limited software is currently available for the Yates-Grundy-Sen estimator, we developed specifications for using the with-replacement sampling variance estimation assumption. We expected a loss in precision, but have found some situations where this is not true. We present a comparison of the estimated sampling variances for the full populations and selected subpopulations for categorical data using both the without-replacement and the with-replacement estimation assumptions. We discuss situations when the with-replacement variances are smaller than the without-replacement variances.

Learning Objectives:

• We will explain to health services researchers reasons for situations when with-replacement sampling variances for a complex survey design are smaller than the sampling variances assuming without-replacement sampling.

Keywords: Statistics, Survey

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.