Making Software Cost Data Available for Meta-Analysis

In this paper we consider the increasing need for meta-analysis within empirical software engineering. However, we also note that a necessary precondition to such forms of analysis is to have both the results in an appropriate format and sufficient contextual information to avoid misleading inferences. We consider the implications in the field of software project effort estimation and show that for a sample of 12 seemingly similar published studies, the results are difficult to compare let alone combine. This is due to different reporting conventions. We argue that a protocol is required and make some suggestions as to what it should contain.