Dollar Unit Sampling: Multinomial Bounds for Total Overstatement and Understatement Errors

摘要: This paper presents a statistical sampling approach based on the multinomial distribution for obtaining bound either total population understatement or overstatement errors both. The is nonparametric in nature and, unlike most currently used techniques, it has known characteristics so that auditor assured of specified confidence level regardless and error pattern. Results are presented which show to give tighter bounds than Stringer all instances studied. behavior studied with respect effects sample size patterns. W HEN auditors first considered use techniques estimating audit amount (or, equivalently, population), they turned classical survey techniques. Inasmuch as have available both book value each unit, was logical them consider ratio difference estimators because these utilize value. [1963] cautioned entails potential dangers their estimated standard will equal zero when no found sample. With large-sample limits this would imply point estimate perfect, which, obviously, not reasonable conclusion an situation employing moderately large