Confidence Intervals for Proportion Estimates in Complex Samples for Performance Audits

• Activity Number: 664 - Making an Impact with Statistical Auditing
• Type: Contributed
• Date/Time: Thursday, August 1, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Statistical Auditing Interest Group
• Author(s): James Ashley* and Carl Barden and Danny Lee
• Companies: Government Accountability Office and Unites States Government Accountability Office and Unites States Government Accountability Office
• Keywords: Assymetric Confidence Intervals; Proportion Estimates; Complex Audit Samples

Abstract

Performance audits routinely require estimation of a large number of attributes to support audit findings, make assessments of internal controls, and assess compliance with laws and regulations. Proportion estimates from these audits are often at or near the boundaries [0,1] and confidence interval estimation requires methods other than normal approximations, especially if they are to be derived from complex sample designs. These methods have been well researched and are now included in several commonly used statistical software packages. This paper describes and analyzes methods applied by the Government Accountability Office to estimate a large number of attribute estimates from a range of complex sample designs. This will include a description of functions developed using three commonly used statistical packages and a discussion of coverage probabilities observed.