ESTIMATION FROM ACCELERATED LIFE TESTS
Abstract
Presented is a statistical technique for analyzing life test data from tests under overstress conditions. Usually the analyst simply assumes that the accelerated and normal use failure distributions belong to specified parametric families, such as the exponential distribution, and that the test experience is a scale transformation of reality. The new technique assumes only that (1) the failure rate is increasing on the average, and (2) test items in the overstress environment tend to fail sooner than those in normal use. Some data from the use environment are required, but may be scanty. Least squares estimators for the life distributions are developed, using both sets of data; quadratic programming may be used for computation. Procedures are given to test the validity of the assumptions used.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1968
- Accession Number
- AD0676261
Entities
People
- Ernest M. Scheuer
- Richard E. Barlow
Organizations
- RAND Corporation