Risk modeling of dependence among project task durations

Project risk management requires assessments of project duration and activity criticality. The assessments, however, can be strongly influenced by the dependence between task durations. In light of the need to address the dependence, the present study proposes a computer simulation model to incorporate and augment NORTA, a method for multivariate random number generation. The proposed model allows arbitrarily specified marginal distributions for task durations (need not be members of the same distribution family) and any desired correlation structure. This level of flexibility is of great practical value when systematic data are not available and planners have to rely on experts’ subjective estimation. The application of the proposed model is demonstrated through scheduling a road pavement project. The proposed model is validated by showing that the sample correlation coefficients between task durations closely match the originally specified ones. Empirical comparisons between the proposed model and two conventional approaches, PERT and conventional simulation (without correlations), are used to illustrate the usefulness of the proposed model.