Level: Intermediate
A questionnaire aimed at improving the current situation needs to ask both result-related and cause-related questions. Then, by grasping the causal relationship between the two through regression analysis, it becomes possible to select targets for which measures should be taken. In this case, missed items cannot be covered after the survey, but if there are any unnecessary results, they can be simply ignored after the fact. For this reason, if all questions are prepared without leaving anything out, with the consideration of not placing too much of a burden on respondents, the result will be a large number of questions and high correlations will appear between items. To address this problem, Takahashi and Kawasaki have proposed "multi-group principal component regression analysis." Its essence is to find principal components for each group based on rational groups that are constructed so that correlation within groups is high and correlation between groups is low, and important principal components are selected using principal component regression with these as explanatory variables, and those with large absolute values of factor loadings on the selected principal components are selected as key items for which measures should be taken.
Sometimes these main items are crowded together, but this can be addressed with factor analysis. When principal components are used in causal analysis, it is called front-side causal analysis, when factors are used it is called back-side causal analysis, and when the two are combined it is called two-sided causal analysis. In this presentation, we will introduce the theory behind this method and a specific method using JMP.
The author has been researching and practicing QM (quality management), SQM (statistical quality management) and design theory for half a century (joint research with numerous companies and management guidance, etc.). Since the 1990s, he has proposed a new design paradigm called Hyper Design, and has researched its mathematical theory, HOPE theory (Hyper Optimization for Prospective Engineering), and has jointly developed its support software, HOPE-Add-in for JMP, with SAS. As a result, he has realized a new design methodology by combining the concept of Hyper Design, the statistical mathematics of HOPE theory, and the support tool, HOPE-Add-in for JMP.
Because design is difficult to approach, it is often misunderstood as a special activity for special people. In order to overcome this and enable more people to master design, the author has developed new educational methods in addition to theoretical research. This is an experiential education method that uses real teaching materials (paper helicopters, paper gliders, coin shooting, etc.) and virtual teaching materials (ball flying simulator, etc.). This educational program (from the basics of statistics to super design) is easy to understand and interesting, as it involves visualized analysis and design using JMP in many situations. This education has been implemented for over 30 years at universities both in Japan and overseas (Keio University, Yale University, Tokyo University of Science, University of Tsukuba, etc.) and at many companies, and its effectiveness has been confirmed.
*For those who wish to understand in more detail, a paper is available. If you are interested, please contact the presenter , Takenori Takahashi, directly.
