Software Engineering Institute

The Causal Discovery and Estimation, novel statistical methods to discover and quantify the causal relationships among variables in a dataset, were applied to the COCOMO and COSYSMO cost estimation models' calibrating datasets.

Results include an identification of which project factors are more likely to affect costs and schedule and a proposed methodology for strengthening the causal signals when handling small datasets.

This research was notable in two regards: (1) USC researchers performed the actual research with coaching from the SEI, who did not have access to the data and (2) Each dataset was relatively small, so null variables were injected into the dataset; and Bootstrapping was employed to reduce false negatives (missed causal relationships) and estimate false positive error (the rate at which spurious causal relationships are detected).