Estimating the R-Indicator, Its Standard Error, and Other Related Statistics with SAS and SUDAAN

• Activity Number: 142
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Michael Witt*+
• Companies: RTI International
• Keywords: Response ; R-Indicator ; Survey Quality ; Weight Adjustments ; Propensity

Abstract

The Representativity Indicators (or R-indicators) are statistics designed to supplement the traditional response rate when evaluating the potential for nonresponse bias from a survey. The response rate provides information on the "amount" of nonresponse in a survey and the R-indicators are designed to measure the similarity between respondents and the original sample or population. In this presentation, we discuss how to estimate the R-indicator using a Logistic Model to estimate response propensity and the Generalized Exponential Model that is used to compute weight adjustments in SUDAAN Release 10. The linearized variance estimate of the R-indicator is presented as well as several related statistics. The variance estimate will account for both the complex design and the model used to estimate response propensity. Example estimates derived using both SAS and SUDAAN will be discussed.