Level: Advanced
【Presenter】
Keio University Graduate School of Health Management
Visiting Professor Takenori Takahashi
[Presentation summary]
Superdesign is an analytical design that uses a superstructure function, which is a dual structure function based on superfactors. This method increases design possibilities and enables advanced designs that were previously impossible. However, the superdesigns developed early on require that the entire interval can be approximated with a single function with high accuracy. However, in recent years, the number of experiments with a large number of levels has increased, and when the number of levels increases from around 10 to more, it becomes difficult to approximate the entire interval with a single function with high accuracy. In such cases, it is reasonable to use a composite approximation in which the entire interval is divided into small intervals and each interval is approximated with high accuracy. However, this means that the number of parameters (coefficients) handled in design increases significantly, so it is necessary to deal with this. This research introduces a method for that purpose. In addition, flexible and sophisticated designs are difficult to achieve with the analysis of variance approach, but are made possible by mathematical programming. What is important at this time is an approach that uses the ``design PDCA cycle''. In this study, we view design as building consensus among the parties involved, and propose a method to achieve this through an approach that involves the "design PDCA cycle" based on mathematical programming. We also discuss flexible design, which allows superfactors to be specified freely, as an advanced application.
[Presenter profile]
For nearly 50 years, I have been conducting research on TQM (Total Quality Management), SQM (Statistical Quality Management), and design theory. Since the beginning of the 21st century, we have proposed a new design paradigm called Hyper Design, developed its mathematical theory HOPE theory (Hyper Optimization for Prospective Engineering), and developed its supporting software HOPE-Add-in with SAS. We are conducting joint development. Through this effort, we have realized a new design method that combines the concept of ``hyperdesign'', the statistical mathematics of ``HOPE theory'', and the support tool ``HOPE-Add-in for JMP''. As a social scientific extension of this theory, he proposed a selective two-sided causal analysis ([multigroup principal component regression analysis] as a front-side causal analysis + [factor regression analysis] as a reverse causal analysis), and is currently using a selective two-sided causal analysis. We propose reverse causal analysis (a causal analysis method equivalent to quasi-SEM), which is an advanced approach after causal analysis.
[Collection of papers]
Go to the collection website
https://community.jmp.com/t5/Discovery-Summit-Japan-2022/Discovery-Summit-Japan-2022-%E8%AB%96%E6%96%87%E9%9B%86/ta-p /568396
*Video will be released until November 30th (Wednesday)
Publication has ended.
