ABSTRACT
A number of statistical models are discussed. Three logistic regression models and a proportional intensity model are applied to software reliability data.
A large software reliability data set was collected which included failure and fault information as well as attributes of the programs such as size, type of program, etc. Execution time had not been collected and so the well known software reliability models were not applicable in this instance. However, the statistical models presented here equate some function of failure count (or time to failure) to the explanatory information available and the results are very encouraging for further research.
The final section shows how the exponential type software reliability models may be incorporated into a proportional hazards framework.
