On selection problems with several non-perfect experts / Nancy Kupfer
In work and private life we have to make decisions every day - knowingly or unknowingly. But in some cases the process of decision making can be very complicated and difficult. For that reason, consulting more than one decider has proved useful in practice. But, are decisions made by several deciders better than decisions made by a single decider in general? In this thesis we investigate six types of selection problems with two or more experts. In each of these selection problems the basic task of the experts is to select k out of n given items with maximal values or deleting n-k out of n items with minimmal values. These experts are not perfect, but the quality of their selected items is better than those of randomly selected items. Because of not allowing any exchange of information between the acting experts, they are totally independent and observe their own preference orders only. Considering two experts A and B with equal noise levels selecting k out of n items, we focus on the question, whether the items selected by A and B are better then the items selected by A only. It shows that each double-expert-scenario with at least one action of expert B is better than letting A select all k items. Moreover, we observed interesting structures within the rankings of all these double-expert-scenarios. Considering experts with different noise levels, we are interested in the best and worst selection orders. For alternately acting experts we achieve the best results if the worst expert selects the first item, the second worst expert the second item, ..., and the best expert the k-th item. To get the worst results, the experts should act in reverse order.
|Dissertation:||Jena, Univ., Diss., 2013|
|Subjects:||Statistische Entscheidungstheorie > Auswahlverfahren|
|Type of content:||Hochschulschrift|
|Related resources:||Erscheint auch als Online-Ausgabe: On selection problems with several non-perfect experts|
|Physical description:||119 S. : Ill., graph. Darst. ; 29,5 cm|
|Basic Classification:||31.73 Mathematische Statistik|