Presentations -

An attempt to estimate micro markets by time series clustering

Kazutomi Sugihara

32nd Hokuriku Branch Symposium and 36th Human Symbiotic System Research Meeting in Japan Society for Fuzzy Thenory and Intelligent Informatics  2023.12  Hokuriku Branch of Japan Society for Fuzzy Thenory and Intelligent Informatics

　More details

This article introduces an example of data analysis using time series clustering as an attempt to estimate "micro markets". To discover it is an essential task in promoting sales at unmanned stores. An example of POS data from vending machines is illustrated to discuss future prospects for estimation.

Attempt to review questionnaire items using item response theory

Kazutomi Sugihara

The 44th Annual Conference of JSET  2024.3  Japan Society for Educational Technology

　More details

Most educational institutions now conduct questionnaires of class evaluation, but if there are flaws in the content or method of the questionnaire, the results obtained may not be accurate. The author considers this situation to be a "dysfunction of questionnaire", and attempts to revise the items using item response theory as a strategy for redesigning the questionnaire using statistical methods. In this presentation, we reported the results of an analysis using item response theory.

Detection of Non-Functional Questionnaire Items Using a Binary Logistic Model

Kazutomi Sugihara

Fuzzy System Symposium 2024  2024.9  Japan Society for Fuzzy Theory and Intelligent Informatics

　More details

In the implementation of various questionnaires, it is conceivable that certain survey items may fail to accurately capture respondents' thoughts. This issue, known as "questionnaire dysfunction," has prompted ongoing efforts to establish a process for reviewing and improving questionnaire items. In this presentation, as part of the questionnaire redesign process, we employ a quantitative approach to detect non-functional items. Specifically, using the binary logistic model within the framework of Item Response Theory (IRT), we estimate item parameters, classify survey items based on their characteristics, and identify candidates for potentially problematic items.

Assessment of Measurement Properties and Development of Improvement Criteria for Questionnaire Items Based on the Two-Parameter Logistic Item Response Model

Kazutomi Sugihara

34th Hokuriku Branch Symposium in Japan Society for Fuzzy Thenory and Intelligent Informatics  2025.12  Hokuriku Branch of Japan Society for Fuzzy Thenory and Intelligent Informatics

　More details

This study examines the measurement properties of questionnaire items using the two-parameter logistic (2PL) item response model and proposes systematic criteria for identifying items requiring improvement. Traditional evaluation approaches often overlook item-level issues such as low discrimination or extreme difficulty, limiting the precision of questionnaire-based assessments. Applying the 2PL model to an existing dataset, this study analyzes the distribution of discrimination(a_j) and difficulty(b_j) parameters to capture the overall measurement structure. Based on these parameter estimates, quantitative thresholds are established to classify items as core items, candidates for revision, or candidates for removal. Qualitative criteria, including ambiguity of wording and double-barreled content, are integrated to complement the statistical diagnostics. The study further illustrates how the proposed framework supports questionnaire redesign, such as the development of short forms. The findings contribute a practical and scalable methodology for enhancing the validity and efficiency of educational and psychological measurement.

Optimization of Questionnaire Scales through Item Revision Based on a Comparison of 2PL and 3PL Models

Kazutomi Sugihara

The 46th Annual Conference of JSET  2026.3  Japan Society for Educational Technology

　More details

This presentation aims to examine the measurement properties of questionnaire items using the two-parameter and three-parameter logistic (2PL/3PL) models of item response theory in order to improve scale accuracy. By comparatively analyzing item discrimination, difficulty, and guessing parameters, we propose a method for identifying items requiring improvement from the perspectives of model fit and information. Through this approach, we seek to systematize criteria for item removal and revision and to develop questionnaire scales with higher reliability and validity.