Level: Beginner
【Presenter】
BioStat Institute Co., Ltd. Representative Director Yukio Takahashi
[Summary of presentation] If you try to analyze two qualitative and quantitative variables in real data, you will inevitably end up with ``repeated ragged two-way layout''. I can't find my textbook. Perhaps for this reason, it is prevalent on the Web that ``it cannot be analyzed unless the repetitions are equal''. The reason for this is that the analysis method using the design matrix X with dummy variables, which relies solely on the decomposition of the sum of squares, is hated in introductory textbooks on statistics. A JMP user using Fit Model should not be concerned about whether the repetitions are equal or uneven. You probably know that when you try to find the level means of a qualitative variable in Effect Details, you get 'Least Squares Mean' and 'Mean', and that they are at odds. But can you explain the difference properly? Therefore, we present an analysis method using Excel's matrix function for 2 x 3 repeated irregular two-way arrangement data, and compare it with the analysis result by JMP to deepen our understanding of the calculation principle of "model fitting". A computational alternative to sum-of-squares decomposition is presented for further applications.
[Presenter profile]
I've been using SAS since JMP4 and before JMP. At the company I used to work for, I conducted statistical education using various types of JMP, and also carried out activities to disseminate JMP within the company. This result has been introduced at JMPer's meetings. After retirement, he established a self-employed consulting company and continues his activities. Recently, I feel that the ability to statistically analyze real data in many academic fields (except for clinical statistics) in Japan is collectively declining. For me, the experimental design method was indispensable for learning statistical theory and practicing it in practice. The theme of this time was a problem that had already been solved for me, but when I read the books related to the design of experiments again, I found that there was no example using the calculation principle of JMP's "model fitting". I was appalled at the parade of traditional analysis methods that insisted on decomposing the sum of squares, and I was convinced that this was a widespread delusion that it was impossible to analyze 2-way data with repeated raggedness.
