Robust Methods: Theory and Application for Construction Projects

Robust statistical procedures are "close" to the optimal parametric procedures when the real distribution coincides with the known one and stably retains its qualities as long as the true distribution is in its vicinity. Just like the qualitative, quantitative approach to determining the robustness of procedures is based on the requirement that arbitrarily small changes in the distribution of observations should cause only sufficiently small changes in the characteristics of the quality of procedures. Application of robust procedures to the assessment of construction costs will reduce labor intensity and increase reliability of estimates. This paper establishes the method for robust parameter estimates of contaminated data sets. The robustness estimation means minimizing asymptotic bias of the estimate in the presence of contaminating observations in the contrast to the approach of P. Huber and J. Tukey, where robustness is minimizing of the asymptotic variance of estimates. Minimax asymptotic bias method, based on the maximum likelihood approach assuming an arbitrary contaminating distribution is developed. Example of applicationis given for construction project concerning the determination efficiency structures of “zero construction cycle”.