OBJECTIVE ? This paper builds on our previous research in which we found inconsistency within and between results in empirical studies of software engineering cost estimation which had compared regression and analogy techniques. To this end, this paper aims to determine why and how these inconsistencies occur. In addition, we attempt to provide a solution which might reduce inconsistencies in future. METHOD ? We retrieved a sample of 11 journal papers and 9 conference papers which compared both regression and analogy-based prediction methods. Analysis of these papers uncovered inconsistencies within and between results. From this starting point we concluded that researchers should look beyond asking ?What is the best prediction system?? Thus we now consider the reasons for these seemingly conflicting results. RESULTS ? We found little consistency between studies both methodologically and in terms of outcome. CONCLUSIONS ? We propose that researchers adopt more rigorous and uniform protocols when empirically investigating project cost prediction methods and interpreting the results.
|Title of host publication||International Conference on Software Engineering Research and Practice (SERP'06), June 2006., Las Vegas, USA|
|Publication status||Published - 1 Jul 2006|