Estimating the total number of software failures using an exponential model
John D. Musa, Anthony Iannino · ACM SIGSOFT Software Engineering Notes · 1991
Different approaches to estimating the total number of failures parameter for the Jelinski-Moranda model are considered. The maximum likelihood estimator, a harmonic jackknife maximum likelihood estimator, and a simple Bayesian estimator are studied. None of these is found to be entirely satisfactory, each suffering from either infinite estimates and/or severe bias in small to moderate samples. The purpose of this paper is to summarize this work, in order to stimulate a search for better estimators.