The Sources and Uses of Sensitivity Information

Publisher Website
Google Scholar
Abstract
The central concept of management science is the concept of a model—that is, a relationship between those variables under the control of a decision-maker (decision variables), those not under his control (environmental variables), and one or more measures of cost or performance. To solve a model means to (1) experiment with the model to calculate the anticipated cost and performance of proposal decisions (simulation) or to calculate the decision variables that minimize or maximize a single measure of cost or performance with constraints on other measures (optimization), and (2) to perform sensitivity analyses that measure the rate of change of the “output” of the model (the cost and performance measures) with respect to the “inputs” (the decisions and the environment). The management science literature has emphasized the former objective, and in many cases the latter has been a byproduct. Optimization techniques often provide some useful sensitivity information with little or no additional computational effort once an optimal solution has been calculated. Simulations do not admit sensitivity calculations as easily. Since the sensitivity studies are usually accomplished by performing multiple simulations with marginally different inputs, the cost of performing such studies can be quite large. Therefore, it appears useful to outline the way in which sensitivity analyses are used in decision-making and to examine the way in which they are generated, with a view to reducing the computational effort needed to produce useful sensitivity information.

This publication has 0 references indexed in Scilit: