Correlation sensitivity analysis in stochastic project networks by Anand

We study the interplay between correlation in stochastic project networks and project completion time. Suppose the activity durations have arbitrary marginal distributions and a Gaussian copula. The impact of an increase in correlation between parallel activities is a stochastic decrease in project duration regardless of network topology, whereas the impact of an increase in correlation between serial activities is dramatically different. We identify an important subclass of serial activities—including all serial activities in series-parallel networks—for which an increase in correlation between serial activities causes an increase in project duration in the increasing convex order sense. When the network is not series-parallel, some serial activities may have an indeterminate effect. We discuss an application of these results to project subcontracting. A special case known to be valid in practice and involving a Gaussian copula is when durations are multivariate lognormal.