A Study on the Application of Referential Hierarchical Clustering Algorithm in the Selection of Questions in Questionnaires

Questionnaires are commonly used in market surveys, polling and academic research. However, there are some problems in the use of questionnaires for survey purposes. Volunteers had to answer too many questions on the questionnaires. As a result, some volunteers would refuse to complete the questionnaire while others might sloppily do so. This phenomenon results in a huge difference between the expected and the actual results of a questionnaire. Hence, there is need to create better questionnaire content - a content which can fully aid in the study of issues that it was designed for in the first place. The proposed method is to apply the Referential Hierarchical Clustering Algorithm (RHA) to analyze results from a survey followed by reordering the questions based on the close relationship between questions. As a result a questionnaire is designed that is more representative of what the survey really needed. Experimental tests were performed by posting questionnaires to students taking the elective Internet course at the College of Literature of the National ChangHua University of Education. Validation tests were made using the Wilcoxon rank-sum test to prove that the questions from the newly created questionnaires are more representative to what the survey needed as in comparison to the original questionnaire. Results showed that the newly created questionnaire had fewer questions and can effectively provide a survey close to the requirements it was designed for. This study can provide scholars a base for selecting questions when using questionnaires in their researches.