- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Report Inappropriate Content
How to display aliased terms in Fit model result
I am dealing with a fractional factorial design,and in the Fit model platform I find that if there is no terms aliased with each other in my model,the Singularity Details dialog will disappear.but I think this is very useful,you can get the alias relationship without further thought.and it is very helpful for me to select suitable terms into my model.is there any setting in jmp that make this dialog always open?
this is the situation happened to me
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Report Inappropriate Content
Re: How to display aliased terms in Fit model result
Hi @ZHANDOUJI,
In your DoE datatable, you can always have a look at the aliasing structure of the design by clicking on the script Evaluate Designs and checking the Alias Matrix and Color Map on Correlations panels.
The Model script in your datatable should also create a model that respects your assumed model and aliasing structure (so you should not see the Singularity Details with the provided analysis script in your DoE table). The aliasing between terms is a consequence of your design (and assumed model), not your (final) model.
In your example, in the first screenshot shows you that interactions A*B and C*D are confounded, so these two terms can't be estimated independantly (hence the missing values for logworth, p-values and biased/zeroed estimates). In the second screenshot, removing one of the terms enable you to estimate the other terms, but taking into consideration that it's not this term alone : in fact you're estimating the estimate for A*B + C*D.
More infos on Singularity Details, Models with Linear Dependencies among Model Terms
Hope this answer will help you,
"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Report Inappropriate Content
Re: How to display aliased terms in Fit model result
Hi @ZHANDOUJI,
In your DoE datatable, you can always have a look at the aliasing structure of the design by clicking on the script Evaluate Designs and checking the Alias Matrix and Color Map on Correlations panels.
The Model script in your datatable should also create a model that respects your assumed model and aliasing structure (so you should not see the Singularity Details with the provided analysis script in your DoE table). The aliasing between terms is a consequence of your design (and assumed model), not your (final) model.
In your example, in the first screenshot shows you that interactions A*B and C*D are confounded, so these two terms can't be estimated independantly (hence the missing values for logworth, p-values and biased/zeroed estimates). In the second screenshot, removing one of the terms enable you to estimate the other terms, but taking into consideration that it's not this term alone : in fact you're estimating the estimate for A*B + C*D.
More infos on Singularity Details, Models with Linear Dependencies among Model Terms
Hope this answer will help you,
"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)