Comparison Of Some Software Reliability Models For Simulated And Real Failure Data
G. Mock · International Journal of Modelling and Simulation · 1984
Modelling and parameter estimation are distinguished as aspects of software reliability evaluation. Three software reliability models have been selected for application to software failure data obtained in practice. The model parameters are estimated by means of the maximum likelihood method. New criteria on the data are derived that have to be satisfied in order that bounded estimates for individual model parameters exist. Application to both simulated and real failure data shows that on the Whole, none of the models performs significantly worse or better than the others. The estimation results are dependent on the amount of failure data available and on the way the data have been obtained.