Lossess representation of questionnaire data

Abstract

Data collected with questionnaires is oftentimes Likert-type. Many research are based on handling and representing this type of data. This thesis introduces a new method to represent Likert-type data, providing a representation that is less prone to information loss. The introduced method is shown to be versatile in case of different types of weights and evaluations. The mathematical base of the method is introduced and it is used on an artificial example to present its behavior. Then it is demonstrated on a real dataset, to show its potential in research. The results are analyzed in relation to the original dataset. Limitations and further development possibilities of the method are discussed.