Dual Kriging: An Exploratory Use in Economic Metamodeling

Sensitivity analysis of capital investments can be effectively carried out by employing a metamodel approach and experimental designs. Although polynomial regression metamodels are popular and straightforward, they do not consider spatial relationships among the data. Dual kriging is an estimation technique that allows the incorporation of spatial correlation into the interpolation or estimation process and has been used primarily in geophysical statistical analysis. This article investigates the dual kriging approach as an alternative technique to polynomial regression and artificial neural networks for metamodel analysis of capital investments. It is observed that dual kriging shows potential in performing sensitivity analysis as its accuracy is as good as, or better than, other techniques, and, while model building and interpretation is more complicated than that of polynomial regression, it is significantly more straightforward than that of neural networks. Furthermore, this is the first known work of using kriging in the field of engineering economics; there may be other useful applications such as in cost estimation.

Notes

¹Magnitude of the standardized regression coefficient.

²Magnitude of the percentage change in the output.

³Magnitude of the standardized drift function coefficient.

¹Magnitude of the standardized regression coefficient.

²Magnitude of the percentage change in the output.

³Magnitude of the standardized drift function coefficient.