Component Variance Methods for Establishing CEP Confidence Bounds Using Component Error Source Statistics.
Abstract
Two methods for calculating upper confidence limits for a variance, Sigma (Subscript T, superscript 2), have been developed for the case where individual contributors to the total variance can be measured. The methods can reduce the upper bound--or the number of tests--considerably when two or more component variances are of roughly equal magnitude. Both methods are conservative (i.e., introduce additional confidence) and yield the classical, chi-square result when a single component variance is much larger than the others. The methods are valid even for cases where the component statistics are based on different numbers of tests and do not require a priori bounds of relative component contributions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1971
- Accession Number
- AD0882321
Entities
People
- Alfred F. Riedel
- John D. Malick
Organizations
- SRI International