Statistical inference of a software reliability model by linear filtering
Mario Hellmich · Journal of Statistics and Management Systems · 2016
We discuss statistical inference of a software reliability model which incorporates the operational profile of the software. In this model the instantaneous software failure rate depends on both the number of remaining faults as well as on the operational state of the software system, which is assumed to change over time according to a Markov process. Statistical inference is performed by means of the theory of linear filtering, using the so- called innovations method and martingale theory. We discuss inference at two different information levels, namely those defined by observing the failure process alone as well as both the failure and operational processes. Software reliability prediction based on the observed data is considered as well, and we provide an explicit formula for software reliability conditional on the data.