Level: Intermediate
【Presenter】
Yukio Takahashi, Representative Director, BioStat Laboratory Co., Ltd.
[Announcement summary]
The traditional design of experiments is not limited to orthogonal arrays, but is an analysis method that mainly uses sigma to decompose the sum of squares on the premise that the factors are orthogonal to each other. However, since this method cannot be applied when missing values occur, it is necessary to perform analysis using the "linear estimation theory" that has provided a theoretical framework for analysis of variance. However, it is difficult to find the book that explains it.
JMP's "Model Fitting" also supports analysis of orthogonal arrays with missing values, but "estimation of the 95% confidence interval of the difference between the current level and the optimum level" is a standard analysis of orthogonal arrays. How should I find it? Is the "Taguchi's formula" or "Ina's formula" for estimating those error variances incorporated in JMP? It is courageous to rely on the JMP results as they are without a book that explains what analysis procedure is used to calculate them.
Therefore, for the case where there is one missing value in the L8 orthogonal array, we show the analysis method using the analytic function of Excel and compare it with the analysis result by JMP to deepen the understanding of the calculation principle of "fitting the model". ， Shows the basic calculation method that can be further applied.
