Statistical comparison between separate-version and short-version questionnaires

Abstract

Short-version questionnaires are often used in psychological research dealing with constructs, because the number of questions that the respondent must answer can be reduced. However, because such questionnaires do not include all of the items of the original scale, it is questionable whether the results have sufficient validity. In this research, a separate-version questionnaire was introduced, which can reduce the number of items in the same way as a short-version questionnaire, but is expected to have a higher degree of validity. A statistical comparison was performed between a short-version questionnaire and the introduced separate-version questionnaire, both by simulation and using real data, where the analysis model was a confirmatory factor analysis. The simulation considered the situation where the data used in a scale development study were different for the short-version and separate-version questionnaires. The results showed that the separate-version questionnaire had smaller biases than the short-version questionnaire, which indicates the usefulness of the separate-version questionnaires.