Choice of Cost-Estimation Functions Based on Statistical Quality Criteria and Technical Coherence

This paper studies the problem of choosing cost estimation functions by mixing statistical regression criteria and technical coherence. The parametric cost estimation method which looks for functional relationship between cost driver variables and the product cost is particularly well adapted to early economic evaluation during the design phase; it is simple to implement, efficient, and reactive enough to provide a powerful guideline for the designer. These cost estimation relationships (CER) are generally computed using linear or nonlinear regression algorithms. However, the problem is to identify and to choose the best CER among several candidate formulae which have been developed. In many cases, the statistical regression quality and estimation criteria are not sufficient. It is also necessary to consider the technical coherence of the CER as defined by the experts in the application domain. Therefore, an additional problem is to make these different points of view consistent. In this paper, the authors adopt a weighted multicriteria decision approach. The characteristics of the different quality criteria are described, then three practical selection methods are presented. All these methods measure the general quality of CERs, taking into account both technical and statistical points of view. The methods are based on a measure of quality using a weighted sum of normalised criteria, a weighted ranking method, and an original use of a similarity measure in the criteria space. These methods are then tested and compared using a real case.