A Bayesian Approach to Extrinsic Versus Intrinsic Uncertainty in Design for Market Systems

This article illustrates how variance in the predictive distribution of the profit objective function in a design for market systems model can be decomposed into two components using a simulation based Bayesian approach introduced in the econometrics literature. The first component, intrinsic uncertainty, would be retained in the model even if the model calibration parameter values, such as parameters representing customer preferences, were known with certainty. The second component, extrinsic uncertainty, stems from lack of precision regarding model calibration parameters such as customer preferences. The simulation based approach overcomes a key problem in decomposing uncertainty for the typical design for market systems problem by overcoming the difficulties associated with analytical treatment of non-normal distributions. The variance decomposition approach is demonstrated for the design of a handheld grinder power tool. Following the same Bayesian decision analysis framework the variance simulation method can be applied to other design for market system problems with other objective functions and with additional sources of uncertainty