This thesis considers a software reliability growth model with modified error detection rate. Maximum likelihood estimators are obtained for the software reliability model parameters, and they are compared with those of Geol-Okumoto and generalized Geol-Okumoto models. To determine when computer software should be released, profit models are given which involve profit from correcting an error found during test phase, cost of correcting an error found during operation phase, and testing cost. Methods of finding the optimal release time are presented and numerical examples are given.