Poisson Models for Subprogram Defect Analyses

For Ada systems, the hierarchy of subprograms compose a layered virtual machine within an object-based framework. Poisson analyses are proposed for the identification of the determinants of defects in these subprograms. Software complexity (measured during design or implementation) as well as characteristics of the software development environment influence the number of defects identified during the testing phase. The Poisson models are calibrated on the basis of measures extracted from the code of Ada projects and data from software change reports. One of the models is used to estimate defects at the subprogram and project levels for the calibration data, and these estimates are then compared to the actual defects. To demonstrate cross-language applicability, defect predictions are made at the subsystem level for a project coded in the C programming language and compared to the actual subsystem defects. Notable results from the analysis are that extensively modified reused subprograms (>25% changed) have substantially more defects than new code of otherwise comparable characteristics and that software development environment volatility (as measured by non-defect changes per thousand source lines of code) is a strong determinant of subprogram defects.