Level: Intermediate
【Presenter】
Yukio Takahashi, CEO, BioStat Laboratory Co., Ltd.
[Announcement summary]
The "least squares mean" and "prediction profile" obtained as a result of JMP's "model fitting" are indispensable for displaying the results of exploratory analysis. However, it is a dialect that is not listed in the index of "Statistics Dictionary" by Takeuchi et al. (1989), and it is difficult to explain it. Recently, we have found that a concise and clear explanation can be made by complementing the "least squares mean" and "predictive profile" with each other in "analysis of covariance" including qualitative factors, quantitative factors and interactions. The data used are the case of "regression analysis including stratified factors" in "Multivariate analysis method, revised edition" by Okuno et al. (1981) (observation result of sulfur dioxide gas recovery tower, concentration of recovery liquid y, washing water The temperature of is x, and the operating period of 4 levels A). In the original work, there is a graph display as if the operation period (January / February) has a different slope than other periods. Using a model that includes interaction, follow the change in the profile of the temperature x of the washing water while changing the operating season A with the prediction profile, compare the average concentration y of the operating season A while changing the temperature x, and compare the temperature x. It can be clearly shown that it matches the minimum squared average of the operating season A when the total average of is given.
https://www.yukms.com/biostat/takahasi2/rec/009-13.htm
[Presenter Profile]
I am a user from JMP4 and have been using SAS since before JMP. At the company I used to work for, I have been conducting statistical education using various JMPs and promoting JMPs within the company. This result has been introduced at the JMPer's meeting as appropriate. After retirement, he established a self-employed consulting company and continues his activities. Recently, I have realized that the ability to analyze statistical data of actual data in many academic fields (excluding clinical statistics) in Japan is declining collectively. Analysis of covariance is an old and new challenge for me. It was necessary to take seriously the analysis of covariance including interactions. Analysis of covariance has become established as an analytical model when there is no interaction. Recognizing that historically it has been discussed to include interactions, this is extremely useful in explaining the least squares dialect of SAS and the predictive profile of the JMP dialect. I was keenly aware of that.
