A macro approach to estimating correlated random variables in engineering production projects

An important consideration in cost risk analysis is the amount of correlation between different cost elements. If correlation is ignored, both the probability and magnitude of costs overruns could be significantly underestimated. The two major difficulties in implementing correlation addressed are estimating correlation coefficients and providing an accurate theoretical risk analysis approach to account for these correlations. Since detailed correlation data are often difficult or impossible to obtain, an intuitive approach is proposed, which estimates correlations for cost estimates relative to several underlying macro factors. The correlation matrix obtained by this method is positive semi-definite and a case study based upon three macro factors is given. The cost risk distributions are computed and compared using the beta fit model and two other more complicated models. This study shows negligible differences in cost risk dollars when computed by the various models. This method of using macro factors to estimate correlation coefficients can account for significant additional cost risk dollars while not requiring external correlation data.