| Abstract
| - Consideration is given to determining when all the faults in a reliability system have been detected, assuming the use of the recapture debugging procedure introduced by Nayak (1988). A stopping rule based on the likelihood ratio is proposed. Compared to the stopping rule suggested by Nayak, this likelihood-based rule makes better use of the available information, and, for a given error level, yields a small reduction in the average time taken to reach a decision. A generalization is suggested for the situation where the faults in the software can be categorized into two or more classes, between which detection rates are permitted to differ.
|
