This paper proposes a systematic approach to certain parameter estimation problems relevant to applied optimization models. A context-specific measure of decision maker performance called decisional efficiency is defined as a function of the unknown parameter vector. The values of this measure are considered to be distributed according to a performance density. Then a principle of Maximum Decisional Efficiency is proposed and its relationship to the Maximum Likelihood principle of statistics is discussed. The principle is first illustrated on a problem of group ideal point estimation. Next, estimation of the ratio of ordering and holding costs in inventories managed by the EOQ model is considered. The results in both these examples are intuitively appealing, suggesting face validity of the principle. A more detailed illustration is developed for the problem of deriving a consensus scoring model for credit applicants which combines the data of more than one loan officer.
