Statistical indices for software quality estimation

Abstract

n major feature is a work package designed to satisfy the processing objectives of a client company. The number of major features always varies from one software release to another C01891. Although there is an increasing trend in the number of major features and enhancement modification requests from one software release to another, the number of maintenance modification requests continues fluctuate; consequently, the total modification requests and items do not always exhibit a stable pattern [01891. In addition to providing information on the quality of discrepancy reports, modification request items, scripts and the overall software, quantitative indices are needed for estimating the discovery rate of discrepancy reports, the defect density of the current software and the field fault density. The primary purpose of this project was to develop statistical indices for estimating the quality of new or updated software releases. This paper presents a set of newly-developed absolute and relative measures that are useful for estimating the quality of software releases. Quality metrics were developed to better support (a) quality assurance estimation for major and minor software releases; and (b) computation of estimated field values such as the field fault density. Uhl ike the existing software quality metrics, (e.g., Li89, MuB7, Hu891, the new sets of quality assurance indices will be found useful by two categories of users — one set for the developers and test organizations, and the other in administrative decision-making positions. Cubic splines, polynomial splines, Langrangian interpolating polynomials and least squares ression analysis are alternative tools