Toward Constructing Confidence Intervals from Business Survey Data: An Empirical Investigation

• Activity Number: 580
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Katherine Jenny Thompson*+
• Companies: U.S. Census Bureau
• Keywords: bootstrap ; confidence interval ; replicate variance

Abstract

Many sample surveys publish confidence intervals (CIs) for key estimates. With a linear estimator and a single-stage design, the literature supports the usage of the t-statistic as the critical value, with degrees of freedom obtained determined by the replication method used for variance estimation or by the stratified survey design, with the possibility of Sattherwaite approximations, or alternatively, the construction of Bootstrap or Bootstrap-t CIs. The theoretical justifications for employing any of these methods assume a linear estimator, and many surveys publish ratio estimates. This paper presents the results of a simulation study that compares the coverage properties of these CIs for a linear estimate and a ratio estimate drawn from a stratified design drawn from a highly skewed population modeled from a business survey, employing several replicate variance estimators.