DOE error DF = 0
Six Sigma – iSixSigma › Forums › Old Forums › General › DOE error DF = 0
 This topic has 4 replies, 4 voices, and was last updated 14 years, 8 months ago by Jen.

AuthorPosts

November 26, 2007 at 4:07 pm #48766
I am trying to analyze a 2factor, 3level full factorial experiment in Minitab but I keep getting a “Denominator of Ftest is zero” message. It also shows that the DF for error is zero. I do not know how this can be and I am at a loss as to how to proceed. I’ve tried multiplying my data by 10 but this does not help. It seems as though no matter what I do to the data I keep getting the same message. Any assistance would be much appreciated.
0November 26, 2007 at 4:23 pm #165325
Robert ButlerParticipant@rbutler Include @rbutler in your post and this person will
be notified via email.If you have a 2 factor 3 level design you will have 9 experiments. If you have told your program you want to test for all of the factors this design will allow this means you are asking the machine to test for
A, B, A*A, B*B, A*B, A*A*B, A*B*B, A*A*B*B and the mean (intercept) – this is 9 terms and you have only 9 data points – you have no degrees of freedom left for error.
If you can’t replicate at least one point in your design you will have to ask your program to run a forward selection with replacement stepwise routine and not a backward elimination in order to come up with a model.0November 26, 2007 at 4:45 pm #165328
Van Kim BanMember@VanKimBan Include @VanKimBan in your post and this person will
be notified via email.First try to understand the DoE concept before using the minitab.
0December 3, 2007 at 5:31 pm #165640REL,
Your DOE design is saturated, you take up all your degrees of freedom with the factors and leave none for the error factor in the ANOVA. You can do a couple of things. First, if you have the resources you can run the experiment again. You will then have a replicate to provide some information for the error variable. Or if any of your settings are the typical settings of what you are analyzing you could turn them into centerpoints and run your analysis with center points. This will give you pvalues to analyze. If the centerpoints show as not being significant you can reduce them from the analysis and continue until you determine if the interaction, a factor, or all are significant. I hope this helps you.0December 3, 2007 at 5:47 pm #165643REL,
Keep in mind what I put in my previous posting will have you reanalyze the data collected as a 2k with centerpoints, you will not have a general full factorial anymore.0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.