|
Activity Number:
|
266
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #302030 |
|
Title:
|
Dynamic Software Reliability Models for Typed Defects
|
|
Author(s):
|
Nalini Ravishanker*+ and Bonnie K. Ray and Zhaohui Liu
|
|
Companies:
|
University of Connecticut and IBM T. J. Watson Research Center and University of Connecticut
|
|
Address:
|
Department of Statistics, Storrs, CT, 06269,
|
|
Keywords:
|
Bayesian inference ; growth curve ; power prior
|
|
Abstract:
|
We describe a model for software reliability characterization of typed defects using a growth curve formulation that allows model parameters to vary as a function of covariate information. For instance, it is possible that an increase in the discovery rate for Type II (assignment/initialization) defects might result as more Type I (checking) defects are found, with possible feedback relationships as well. In general, the dependence of discovery of defects of Type I on an arbitrary number of other defect types can be handled in a dynamic growth curve framework by modeling the reliability growth of Type I defects using covariates to represent the number of defects of other types found up to a certain time. We describe a Bayesian framework for inference and model assessment, using Markov chain Monte Carlo techniques, which allows for incorporation of historical information and expert opinion in the form of prior distributions on the parameters. The use of power priors to combine data-based and qualitative prior information is also introduced in the growth curve modeling context. The methods are illustrated using simulated and real defect data.
|
