Dr. Frank Yang 问答专帖

思源致远

brainchip 发表于 2013-11-5 21:34
杨老师您好，我在学习中又有了一个问题要向您请教。如下图所示，我在做因子筛选的时候，模型的有效性检验是 ...

请将Jmp数据表附上，写明数据的性质，比如说是试验设计的数据，什么样的设计，或者是非试验设计数据，如随机收集的数据。针对性的回答可以避免可能的差错。

思源致远

brainchip 发表于 2013-11-5 22:18
另外再问您一个问题，在JMP的analyze菜单下，fit model和modeling子菜单中的screening命令有什么区别么？谢 ...

After creating and viewing the data table, you can analyze the data in the Screening or Fit Model platforms.
• If your factors are all two-level and orthogonal, then all of the statistics in the Screening platform should work well.
• For highly supersaturated main effect designs, the Screening platform is effective in selecting factors, but is not as effective at estimating the error or the significance. The Monte Carlo simulation to produce p-values uses assumptions that are not valid for this case.
• If you have categorical terms with more than two levels, then the Screening platform is not appropriate for the design. JMP treats the level numbers as a continuous regressor. The variation across the factor is scattered across main and polynomial effects for that term.
• If your data are not orthogonal, then the constructed estimates are different from standard regression estimates. JMP can pick out big effects, but it does not effectively test each effect. This is because later effects are artificially orthogonalized, making earlier effects look more significant.
• The Screening platform is not appropriate for mixture designs.

思源致远

Comparing Screening is different from Fit Model in the following aspects:
Estimates labeled Contrast. Effects whose individual p-value is less than 0.1 are highlighted.
• A t-ratio is calculated using Lenth’s PSE (pseudo-standard error). The PSE is shown below the Half Normal Plot.
• Both individual and simultaneous p-values are shown. Those that are less than 0.05 are shown with an asterisk.
• A Half Normal plot enables you to quickly examine the effects. Effects initially highlighted in the effects list are also labeled in this plot.
• Buttons at the bottom of the report also operate on the highlighted variables. The Make Model button opens the Fit Model window using the current highlighted factors. The Run Model button runs the model immediately.
There is an example to illustrate these in JMP DOE User Guide manual.

brainchip

思源致远 发表于 2013-11-6 20:54
请将Jmp数据表附上，写明数据的性质，比如说是试验设计的数据，什么样的设计，或者是非试验设计数据，如 ...

杨老师，我将数据表附在39楼的原帖上了，这个数据表用的是plackett-burman设计，N=12，是试验设计的数据。我对数据的分析用的是fit model，考虑到是plackett-burman，因此仅选取了主效应进行分析，应用的是最小二乘法，谢谢您！

brainchip

思源致远 发表于 2013-11-6 21:04
Comparing Screening is different from Fit Model in the following aspects:
Estimates labeled Contra ...

Got it. I remembered that I have read this part before but just forgot it. Thanks.

思源致远

brainchip 发表于 2013-11-6 21:54
杨老师，我将数据表附在39楼的原帖上了，这个数据表用的是plackett-burman设计，N=12，是试验设计的数据 ...

你的数据只有3个自变量，好像不对。请检查一下再发一次。

brainchip

思源致远 发表于 2013-11-8 11:53
你的数据只有3个自变量，好像不对。请检查一下再发一次。

杨老师，我上次确实发错了，现在附上数据再发给您，谢谢您！

思源致远

你的数据是Plackett-Burman non-regular 设计，过去的分析仅考虑main effects, 现代的分析方法是考虑main effects+all 2-factor interactions 的stepwise regression method.  If you use the default stepwise regression, you will get

思源致远

根据p value, 你可以将B和A*B从模型中剔除， 得到如下结果：

brainchip

思源致远 发表于 2013-11-8 22:38
根据p value, 你可以将B和A*B从模型中剔除， 得到如下结果：

谢谢杨老师的详细解答！祝您周末愉快！