Directional Pattern based Clustering for Quantitative Survey Data: Method and Application

Abstract

Analysis of survey data is a matter of significant concern as it plays a key role in organizational and behavioral research. Quantitative survey data possesses several distinct characteristics i.e., fixed small range of ordinal values, importance of respondent category labels etc. Due to such reasons quantitative survey data is not appropriate for existing analysis methods involving aggregate statistics. Literature has advised to utilize pattern based analysis tools instead of aggregate statistics since patterns are more informative and efficient in reflecting respondents’ preferences. Thus, we introduce a specialized pattern based clustering technique for survey data that uses the convention of direction instead of magnitude. Further, it does not require manual setting of clustering parameters whereas it automatically identifies respondent categories and their representative features with the help of an adaptive procedure. We apply proposed method over an original academic survey dataset and compare its results with K-Means clustering method in terms of interpretability and usability. We utilize benchmark stakeholder theory to verify the results. Results suggest that proposed pattern clustering method performs far better in segregating survey responses according to the stakeholder theory and the clusters made by it are much more meaningful. Hence, results empirically validates that pattern based analysis methods are more suitable for analyzing quantitative survey data.