Μοντέλα σύγκρισης αξιολογήσεων και προτιμήσεων

Master Thesis
Author
Παπαμιχαλάκης, Χαράλαμπος Γ.
Date
2011-05-10View/ Open
Abstract
The method of paired comparisons has great practical simplicity. It has been used extensively in experimental situations where the subjective judgments or appraisals of individuals lead to qualitative comparative responses, situations where quantification through measurement is difficult or illusory. The method has led to a surprising amount of model building to provide stochastic representation of the experimental process and simultaneously a simple illustration of the role of statistics in science. Searching in bibliography is an easy way to find a considerable number of sources which include relevant studies. The great interest paired comparisons present, has had and nowadays still has resulted in the accomplishment of a substantial work concerning this particular field. Sedulous and detailed examination of the method led the researchers to: a) the stochastic modeling of situations involving paired comparisons, b) the experimental design for paired comparisons and c) the analysis of data deriving from such comparisons. The main part of this project is an interpretative representation of the basic Bradley-Terry model for paired comparisons. The first chapter includes the analytical representation of the model, its evolution until now, a short flashback and its application fields. In the context of the second chapter our study focuses on the model’s numerous generalizations, while the third chapter deals with the representation of some alternative models and the description of the connection method between them and the Bradley-Terry model. The fourth and last chapter encompasses the examination and comment of the model’s fitting capability through the analysis of characteristic applications using statistical packages SPSS, S-Plus and programming language R.