A Comparison of Software Quality Modeling Techniques.
Justin M. Beaver, Guy A. Schiavone · Journal of International Crisis and Risk Communication Research · 2003
Accurately representing the quality of software under development remains a major challenge to the software engineering discipline. Many approaches have been offered that attempt to capture the complex relationships between metrics captured during design, and the final quality of the software product Unfortunately, most of these approaches are either controversial in terms of their validity or simply not universally applicable. This study compares the effectiveness of three different estimation techniques: least square regression, relative least squares regression, and the iterative averaging algorithm. Using the average relative error criterion, it is the iterative averaging algorithm, a very simple method of spatial data analysis, that is revealed to be superior in terms its ability to correctly represent a given data set, and its ability to provide accurate and valid predictions about whether or not a given software module will be error-prone.