Probabilistic Cost Estimating with Subjective Correlations

Abstract

One of the difficulties in probabilistic estimating is accounting for the existing correlations among cost components modeled as random variables. Even if the estimator is aware of the existence of correlations among random variables, calculating accurate values of correlation coefficients is not feasible most of the time. A methodology for generating correlated random numbers in a Monte Carlo simulation for construction cost estimating is reviewed. A methodology is then suggested that simplifies the process of incorporating the effect of correlation coefficients in probabilistic estimating. This methodology consists of assigning subjective measures of correlation between variables. Also, a method is suggested for adjusting the covariance matrix in which correlation estimates are not accurate. Procedures presented are further explained by two examples using actual construction cost data.