AN EXPERIMENT IN SOFTWARE SIZING WITH STRUCTURED ANALYSIS METRICS

This paper presents an experiment that was carried out to develop and statistically validate a model that estimates the number of lines of code in a system. The experiment, described using an experimentation framework, was conducted at three Canadian organizations. A linear model is developed for each of the two samples studied. These models are based on measurements taken on data flow diagrams. Statistical analysis reveals that the models produce sound estimates of the actual data (R2 = 0.724 and R2 = 0.563). Four evaluation criteria show that these models compare favorably to the other software-sizing models found in the literature.