Mixing Combination and the Acceptance test
Abstract
When all of the appropriate hypotheses apply, combinations of 95% confidence ellipses produce resultant ellipses which are themselves 95% ellipses. In other reports it has been shown that when one or more of the hypotheses is violated (such as using estimated covariance matrices instead of the true ones) the result need not be a 95% ellipse. The question addressed in this report is how applying acceptance tests before combination affects the confidence level of the resultant ellipse. The discussions in this document are organized as follows: Section II - A model of the way combination and testing mix; Section III - A discussion of results and conclusions; Section IV - Analysis of limit cases of extreme relative size; Section V - Discussion of when shape can and cannot be ignored, with an example; (Orientation is also discussed). Section VI - Simulation results plus a discussion of the meaning of the values computed. Keywords: Confidence levels; Fix estimation; Chi square statistics; Error ellipses.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 07, 1986
- Accession Number
- ADA196499
Entities
People
- Michael W. Rennie
Organizations
- California Institute of Technology