Using empirical knowledge for software process simulation: A practical example

This thesis describes the combination of real experiments with software process simulation. It introduces different types of real experiments and simulations and describes three main possibilities of combining them: using empirical knowledge for simulation, using simulation for real experiments, and online simulations. The first possibility is explored further by the systematical development and calibration of an executable discreteevent simulation model of a requirements inspection process. The model supports an arbitrary number of documents, each inspected by an arbitrary number of reviewers. The model was developed systematically by following an existing method [RNM03]. The following steps were conducted: the creation of a static process model, the collection and analysis of empirical data, the subsequent creation of an influence diagram, and the final creation and calibration of the dynamic model. The model was calibrated with data from a requirements inspection experiment and two replications. Furthermore, the thesis explains model limitations and lessons learned, and gives a detailed description of how to use the model. A first model validation with real-world data that was not used for the model development suggests that the model indeed reflects reality decently: The deviation between simulation results and real-world data averages 3.3 percent. Conclusions on possible further usage of the simulation model and modeling results complete this work.