Comparison of Results Obtained from Hierarchical Cluster Analysis Methods and Different Distance-Similarity Measures in Determining the Factor Structures of Scales

Abstract

The general purpose of this study is to compare the results obtained by using different distance and similarity measures in hierarchical cluster methods that can be used to reveal the factor structures of the scales. The anxiety scale developed by Büyüköztürk (1997) and applied to 954 university students studying at Muş Alparslan University in 2012 was used within the scope of the research. SPSS and Lisrel package programs were used in the analysis of the data. As a result of the analyzes, the methods that give the best factor structure in Hierarchical cluster methods used within the scope of the research are the Average linkage and the Ward method. When the calculated distance and similarity measures are compared, the cluster results obtained by using Euclidean, Squared Euclidean, Minkowski, Manhattan City Block, Pearson, and Cosine methods are similar in all Hierarchical methods in general. However, it was concluded that the measures that gave the best factor structure in all methods used within the scope of the research were Pearson and Cosine similarity measures. When researchers perform cluster analysis on any educational continuous data, it will be practical in terms of time and effort to first prefer the Pearson and Cosine method and the Average linkage and Ward’s method in hierarchical clustering methods, to obtain the best result.