Quantification of categorical data: comparison of approaches

<p>Measurement problem is one of the most basic in the humanities, since the measurement in this area, which is based on information obtained from a person and as a rule, does not presuppose specific unit as a reference, which would serve as a starting point for comparison. Measurement task is to prove that the numbers system uses corresponds to the empirical area of study, and reflects it adequately. Success in solving this problem is one of the key points in determining the quality of sociological research.</p><p>It is assumed that in a continuous line of consciousness respondent placed his true beliefs about the objects, but when I select an answer from the existing options sign is divided into intervals. When analyzing the results obtained should be considered that this estimate is still inaccurate for psychological personality characteristics of different respondents.</p><p>Quantification techniques allow us to obtain a true representation of the interests of the respondent researchers aim to reach the psychological continuum in the respondent&#39;s mind.</p><p>Each of these methods is based on certain theoretical and mathematical models, which should be taken into account when applying these methods in practice.</p><p>The aim of this work is to compare these methods in terms of application in the human dimension, to develop recommendations for their practical application in the humanities, i.e. comparison of methods of digitizing categorical attributes in terms of application in sociology and develop practical recommendations for their use in sociology.</p><p>Relevance of this study is due to the fact that the method of digitizing data allow to obtain the values of dial implicit (latent) characteristics under consideration behind categorical. This will in addition to the methods used for categorical data, use a combination of methods used in the analysis interval traits make their interpretation more accurate and complete, and as a result, improve the quality of the survey.</p><p>In order to achieve this goal, the following objectives are reached:</p><p>1. Analyze and compare the theoretical and methodological foundation applicability investigated methods of digitizing and methods behind the mathematical model;</p><p>2. Based on comparing the techniques to further select the application to real data;</p><p>3. Apply methods to real data and compare the results;</p><p>4. Develop recommendations for the application of methods in relation to sociological problems.</p><p>The first chapter is devoted to a brief review of the theory of measurement and its role in sociology and modeling process. The second chapter is a theoretical study of methods for categorical data digitization. The third chapter presents the results of applying the methods to capture the empirical data, the results of the use of cluster analysis to digital data and provides some guidance on the application of methods of digitization. Results of the study clearly presented charts and tables both in the chapter and in the appendix.</p><p>The study identified the criteria necessary for the implementation of digitization:</p><p>First, optimization methods require two variables related to each other and are displaying true latent variable on different planes for digitization based on a predetermined system of labels, only one variable, since it is only a projection of the desired latent variable;</p><p>Second, as a result of the optimization, it is possible to obtain different digitizing barchart digitized values of variables. It depends on the variables which are initially selected for digitization and their quantity, and due to the fact that the values obtained are the projections of the latent variable on different planes.</p><p>Despite the fact that the aim of both is to provide methods considered latent variable values, the results of digitizing get different, as evidenced by the different accuracy obtained on the basis of their clustering. Optimization method of digitizing allowed a more accurate subdivision respondents into clusters.</p><p>Based on t