Research synthesis in software engineering: a case for meta-analysis

W. Hayes
Proceedings Sixth International Software Metrics Symposium (Cat. No.PR00403)  
The use of meta-analytic techniques to summarize empirical software engineering research results is illustrated using a set of 5 published experiments from the literature. The intent of the analysis is to guide future work in this area through objective summarization of the literature to date. A focus on effect magnitude, in addition to statistical significance is championed, and the reader is provided with an illustration of simple methods for computing effect magnitudes. Background The field
more » ... f empirical software engineering research can benefit from research synthesis techniques that help summarize and assess the body of empirical results accumulating in the literature. Research synthesis techniques that go beyond the subjective review of literature found in nearly every dissertation could help researchers to build on a framework of existing results, rather than merely commenting on it. In order for the benefits of research synthesis techniques to be realized however, basic standards of research conduct and reporting must be present. Failing some level of comparability among subject populations, sampling standards, and research designs, summaries of existing research tend to focus on the shortcomings or strengths of the research methods used -rather than the results and data reported. Meta-analytic methods allow us to summarize the outcomes of previous research, and form empirically derived expectations for future research focus.
doi:10.1109/metric.1999.809735 dblp:conf/metrics/Hayes99 fatcat:4qkqfteplzephjcd6izqspxjti