DOE
Six Sigma – iSixSigma › Forums › Old Forums › General › DOE
- This topic has 2 replies, 3 voices, and was last updated 17 years, 7 months ago by
DD.
-
AuthorPosts
-
October 7, 2004 at 11:17 pm #37142
Hi, I need your help to analyse the meaning of this information given by minitab for a DOE project:
I understand nothing its important due we have a P value bigger than 0.05 in all factors, but I dont know if my conclusion is right. We have an error of 3%, so the model is complete.
************************************************************************General Linear Model: Yield versus Layers, Method, Blocks
Factor Type Levels Values
Layers fixed 4 0 1 2 3
Method fixed 2 1 2
Blocks fixed 3 1 2 3
Analysis of Variance for Yield, using Adjusted SS for Tests
Source DF Seq SS Adj SS Adj MS F P
Layers 3 0.0008050 0.0008050 0.0002683 1.89 0.249
Method 1 0.0000707 0.0000566 0.0000566 0.40 0.555
Blocks 2 0.0001656 0.0001439 0.0000719 0.51 0.630
Layers*Method 3 0.0000158 0.0000116 0.0000039 0.03 0.993
Layers*Blocks 6 0.0003949 0.0003841 0.0000640 0.45 0.819
Method*Blocks 2 0.0002023 0.0002023 0.0001012 0.71 0.534
Error 5 0.0007085 0.0007085 0.0001417
Total 22 0.0023627Unusual Observations for Yield
Obs Yield Fit SE Fit Residual St Resid
4 0.99530 0.99530 0.01190 -0.00000 * XX denotes an observation whose X value gives it large influence.
0October 8, 2004 at 12:33 am #108718
Robert ButlerParticipant@rbutlerInclude @rbutler in your post and this person will
be notified via email.There are a number of things you might want to consider before concluding you have no significant effects.
1. Your response is yield. If all of your data is in the form of 99.XXX then what you are really looking for are factors impacting XXX. drop the 99 and rerun the regression on X.XX and see if you get the same result.
2. Are your factors discrete or continuous? If they are discrete, did you code them properly for use in the regression package? If they are continuous why haven’t you checked at least for the squared effects of all three? Did you set up the design to check squared x whatever interactions? If you did, where are these terms?
3. With the existing model did you check your plots of residuals against predicted as well as against all of the X’s? If you didn’t you should – you may find some surprising plots that might point you in the direction of a solution.
4. You say “We have an error of 3%, so the model is complete” – what does that mean? A 3% error, if all measures are 99.XXX , sounds like it might exceed the entire range of the response.
0October 8, 2004 at 6:56 am #108721Hi GB,
Apart from what robert has suggested please check your Measurement system. Have you done Gage R&R/bias/accuracy/linearity/stability whichever is applicable. I feel there could ba a Gage issue somewhere in your measurement.
DD0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.