Determining the Appropriate Implementation of a Repeated Measures Analysis Comparing Two Periodic Estimates on Economic Survey Data

• Activity Number: 89
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Philip Steel*+ and Paul Ferrara and Katherine Jenny Thompson
• Companies: U.S. Census Bureau and University of Virginia and U.S. Census Bureau
• Keywords: Economic Indicator ; Hypothesis Testing ; Simulated Data ; Repeated Measures

Abstract

Several of the leading economic indicator programs at the U.S. Census Bureau provide estimates of monthly or quarterly change. Different implementations are being used to test these change estimates for significant difference. Test implementation is derived from the form of the null hypothesis used for the test. The goal of the presented research project is to investigate the statistical properties of these alternative implementations over repeated samples. We consider three variations: a direct comparison of the ratio of two concurrent estimates to 1; a repeated measures analysis of the estimates' difference using point estimates; and a repeated measures analysis that uses smoothed estimates of variance and autocorrelation in the construction of the test statistic. Our research uses two simulated population datasets modeled on the Monthly Retail Trade Survey.