|
Activity Number:
|
49
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 8, 2004 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Quality and Productivity
|
| Abstract - #302114 |
|
Title:
|
A Statistical Test for Compatibility of Two Software Usage Environments
|
|
Author(s):
|
Daniel R. Jeske*+
|
|
Companies:
|
University of California, Riverside
|
|
Address:
|
Dept. of Statistics, Riverside, CA, 92521,
|
|
Keywords:
|
software reliability ; score test ; operational profile
|
|
Abstract:
|
An estimate of the user-perceived failure rate of a software system is typically obtained by extrapolating an analysis of failure data collected during the test and debug phase. An important underlying assumption that validates the extrapolation is that the software was tested in a manner that emulates how the users will interact with it. Testing of this type is referred to as operational profile testing. Operational profile testing is difficult to do in practice, and consequently the extrapolated failure rates frequently have limited usefulness. A partial score test is developed for testing the hypothesis that the test and user environments are identical. Application of the test will provide an early indication in the user phase that the that failure rates extrapolated from the test phase are not valid. It is shown how the failure rates can be recalibrated if the hypothesis is rejected. The size and power of the hypothesis test are discussed. The underlying software reliability model is assumed to be the nonhomogeneous Poisson process proposed by Goel and Okumoto. Extensions to other nonhomogeneous Poisson processes are discussed.
|
