Isolating and Predicting Risks in Architectural Design - Metric Analysis Replication Data

To examine the relationship between container level design metrics and risk, a non-experimental quantitative approach that combined causal comparative and correlation design was selected. A causal comparative approach was needed to test the relative performance of the different risk container types, and a correlation design was needed to test the strength of association between the container design metrics calculated for each container type and error-proneness/change propagation.

The causal comparative correlation testing required sources of comparable design and outcome data from software development projects. Outcome data for the maintainability risks selected for testing, error-proneness and change propagation, is collected from source code repositories such as Subversion and Git, and issue tracking systems such as Jira. Section 3.2 of the thesis further expands upon the details of, and rationale behind, the specific project selection criteria as well as explaining how the selected projects meet those criteria. This replication data set contains the data collected for the four projects selected.