In a typical survey on the outcomes of a lecture event, we did not conduct a statistical evaluation of latent variables said to influence responses to the questionnaire items (observed variables). Instead, we evaluated actual survey results using a formative model that explains the statistically predetermined characteristics of latent variables based on the observed variables. Furthermore, we set latent target items for the questionnaire and examined, using the preference set-based design method, what kind of questionnaire would best fulfill these targets. As a result, based on Structural Equation Modeling (SEM), we demonstrated that it is possible to extract truly meaningful latent items from a concrete questionnaire, and to derive the appropriate response levels for each questionnaire item (observed variable) that would achieve the predefined target levels of those latent items using the preference set-based design method.
ISHIKAWA et al. (Wed,) studied this question.