Home › Forums › Old Forums › General › DOE ANOVA in Minitab
This topic contains 8 replies, has 5 voices, and was last updated by Craig 9 years ago.
I was doing a 2-level, 5-factor fractional factorial DOE with Minitab. In the “create Factorial design” part, I selected 5 factors and a resolution V and a half factorial. In “Analyze factorial design’, I selected the response factor. At the end of the analysis I got the folowing table:Analysis of Variance for resp (coded units)Source DF Seq SS Adj SS Adj MS F P
Main Effects 5 13.60 13.60 2.721 * *
2-Way Interactions 10 41.20 41.20 4.120 * *
Residual Error 0 * * *
Total 15 54.80
What should I do to have Minitab show the Main effects of the model seperatly so that I can guage the significance of each factor?
I was doing a 2-level, 5-factor fractional factorial DOE with Minitab. In the “create Factorial design” part, I selected 5 factors and a resolution V and a half factorial. In “Analyze factorial design’, I selected the response factor. At the end of the analysis I got the folowing table:Analysis of Variance for resp (coded units)
Source DF Seq SS Adj SS Adj MS F PMain Effects 5 13.60 13.60 2.721 * *2-Way Interactions 10 41.20 41.20 4.120 * *Residual Error 0 * * *Total 15 54.80
What should I do to have Minitab show the Main effects of the model seperatly so that I can guage the significance of each factor?
Hello, Rafael, and Happy Holidays…and Happy New Year to you!
It’s a bit difficult to be sure what the problem is without reproducing your problem on my own computer, but here is my best guess as to how to get the information you want:
After you pull down the menu for “Analyze Factorial Design”, click on the button marked “Terms”. Then, where it says “include terms in the model up to order….”, select 1. After you click “OK”, you should get a table with just main effects. You can then “edit last dialog” and try order of 2 to see it with main effects and 2-way interactions.
Then, you can reduce the model, step by step, removing the least significant two-way interactions first until you have reduced the model to just the terms that meet your criteria for significance in terms of their p-values.
Best regards,Eric Maass
I don’t know Minitab so I can’t help you with the specifics of the program but you should be aware that you have zero degrees of freedom for error so with the approach you are taking you won’t be able to assess much of anything.
It would appear you took the half rep of a 2**5 and didn’t bother to add an additional run for a replication of one of the 16 points in the design. Under these cicrumstances you will have to run a stepwise (forward selection with replacement) analysis of the data in order to build up a collection of non-significant terms whose degrees of freedom can then be used for an estimate of error. The regression approach will also show you each effect separately so you can gauge the significance of the term.
Robert,
What is the best way to contact you?
Regards,
Bill
Thank you guys for your help. I tried Eric’s method and it took care of the degree of freedom for the error, this is the table that I obtained when I changed the term to ‘1’Analysis of Variance for resp (coded units)Source DF Seq SS Adj SS Adj MS F PMain Effects 5 13.60 13.60 2.721 0.66 0.662Residual Error 10 41.20 41.20 4.120Total 15 54.80Thank you Bill,
My email address is habib962002@yahoo.com
I suppose you could say that you have solved the degrees of freedom issue for error by using this approach but I don’t think it is a particularly satisfactory solution. If you take a look at your two tables all you have done is arbitrarily declared all two way interactions to be error terms – notice the df, SS Adj SS, etc. for the error are identical to those of the two way in your earlier post.
My understanding of Eric’s post was that he was recommending a modification of the backward elimination procedure which would work if you had df for error. Since you don’t the only way I know to check for term significance (given that you want to look at interactions and main effects) is to run a forward selection with replacement.
The other thing I find disturbing (this may be nothing more than the fact that I don’t know Minitab) is the ANOVA table. You ran a half rep of a 2**5. This presumes 5 factors. Every ANOVA table I’ve ever generated would list the 5 factors and give me a P value for each one instead of lumping them all together and giving me a P value for the Main Effects as though they were just a single factor with 5 degrees of freedom.
That is exactly what I was trying to do. Having the main effects summed up together do not tell me much. I am still trying. Thank you for your help.
Rafael,
To get the individual terms, you can try this:
Select Stat > ANOVA > Balanced Anova. You will see a dialogue box that will allow you to specify terms. The help option within this dialogue box is pretty self-explanatory. Enter in interaction terms by using the asterisk. (A*B).
Is this what you are looking for? You will need some background knowledge on balanced versus unbalanced designs, crossed versus nested terms; fixed, random, and mixed effects models.
HACL
The forum ‘General’ is closed to new topics and replies.
© Copyright iSixSigma 2000-2017. User Agreement. Any reproduction or other use of content without the express written consent of iSixSigma is prohibited. More »