Citation

Gubitz M, Ott KO. Reliab. Eng. 1983; 5(3): 157-171.

Copyright

(Copyright © 1983, Elsevier Publishing)

DOI

10.1016/0143-8174(83)90011-2

PMID

unavailable

Abstract

A method based on isotonic regression analysis of the software failure statistics is presented. The basic information needed for this analysis are the execution times between failures during the test period. The method allows an evaluation of the software reliability in terms of the combined rate of the residual potential failures for which a statistical upper limit is obtained. It also gives indications of the extent to which further testing may be rewarding and a rough estimate of the time needed for further testing in order to achieve some set reliability level.

The Isotonic Regression Analysis (IRA) method has been applied to three examples: the testing of a single module, of a system comprised of a number of modules, and of the practical application of a system, in an operational environment. The analysis is completed with null hypothesis tests of the statistical significance of the improvement of the software reliability indicated by the IRA procedure.