Statistical dependence in risk analysis for project networks using Monte Carlo methods

Monte Carlo simulation of project networks is increasingly used by engineering firms to analyze schedule/cost risk for bidding purposes. However, one serious methodological flaw of most Monte Carlo simulations is the assumption of statistical independence of activity durations in the network. In this paper, a method is proposed to model and quantify positive dependence between uncertainty distributions of activities. This method inherits the theoretically sound foundations of the rank correlation method, but provides a less cumbersome method to elicit dependency information from project engineers. Details of the methodology are described along with an example of project risk analysis in a manufacturing domain (shipbuilding).