Bought the book and read the SPC/CP portions when I was exploring that subject. Here I am creating a DOE and forgot the book goes through that subject as well. I had stopped after I got what I needed with the SPC material. Shame on me but can't wait to complete it now.
This strategy of starting with a full or fractional factorial and using centre points at the beginning, middle and end of the experiment to check that the design space is stable and to check for curvature (then augment with second order model if curvature is confirmed) is just fantastic. Understanding this video is a real breakthrough for me in terms of having an effective sequential approach to experimentation / problem solving. Thanks Paul (Mark T)
Mark - Using the centre points through the DOE is new to me as well. I came across it a year ago at one of my clients. Just shows there is always more to learn....
Hi Paul, great video. How does Minitab use centre points in the factorial response optimiser? We did centre points and used to detect curvature and for target setting of distance. But we noticed that the response optimiser did not use centre point data for calculatons for those who did not import data into Minitab. Would not have noticed except that one guy was entering data into Minitab and then sent the data to others. When they imported the data and noticed they got different target settings for the same distance. When investigating, the ones who imported data accidently forced Minitab to use the centre point data for its reponse model. Not sure if you have seen the same thing? It is not explicitly mentioned on Minitab site that for the final optimization model that the centre point data will not be used in target calculations.....let me know your thoughts. Thx Roger
I have done some experiment with 3 factors 2 steps each and a center Point in the middle of the space. How do i proof the significance of the center Point in a statistical way?
Schwini - Good to hear from you. Simple way is the remove the centre points from the data. Analyse the data as a 2 level DOE then see it predicts the centre points or not. If it misses the centre point values then the centrepoints are needed in the model. If the 2 level model hits the mid-points, then there is no need for a 3 level model... Send me the data paul.allen@allenp.co.uk if you're still unsure...
@@paulallen5321 it was of course linear. It was an academic example just to get comfortable with DoE. So i didnt expect any curved-nature in the first place. :D Thanks for your superb explanations and support!
Bought the book and read the SPC/CP portions when I was exploring that subject.
Here I am creating a DOE and forgot the book goes through that subject as well. I had stopped after I got what I needed with the SPC material. Shame on me but can't wait to complete it now.
This strategy of starting with a full or fractional factorial and using centre points at the beginning, middle and end of the experiment to check that the design space is stable and to check for curvature (then augment with second order model if curvature is confirmed) is just fantastic. Understanding this video is a real breakthrough for me in terms of having an effective sequential approach to experimentation / problem solving. Thanks Paul (Mark T)
Mark - Using the centre points through the DOE is new to me as well. I came across it a year ago at one of my clients. Just shows there is always more to learn....
This was really helpful in my experiment. Thank you.
Truly fantastic! Thank you.
Thanks - can I ask a question....do you use DOE in your motorsport work?
Hi Paul, great video. How does Minitab use centre points in the factorial response optimiser? We did centre points and used to detect curvature and for target setting of distance. But we noticed that the response optimiser did not use centre point data for calculatons for those who did not import data into Minitab. Would not have noticed except that one guy was entering data into Minitab and then sent the data to others. When they imported the data and noticed they got different target settings for the same distance. When investigating, the ones who imported data accidently forced Minitab to use the centre point data for its reponse model. Not sure if you have seen the same thing? It is not explicitly mentioned on Minitab site that for the final optimization model that the centre point data will not be used in target calculations.....let me know your thoughts. Thx Roger
Roger - I think the answer to your question is to use a DOE from the RSM menu, there you can enter centre point data and create curved models..
RSM = response surface
I have done some experiment with 3 factors 2 steps each and a center Point in the middle of the space. How do i proof the significance of the center Point in a statistical way?
Schwini - Good to hear from you. Simple way is the remove the centre points from the data. Analyse the data as a 2 level DOE then see it predicts the centre points or not. If it misses the centre point values then the centrepoints are needed in the model. If the 2 level model hits the mid-points, then there is no need for a 3 level model...
Send me the data paul.allen@allenp.co.uk if you're still unsure...
@@paulallen5321 thank you very much for your support. I just simply plottet my data and the centerpoint to prov the linearity of my data.
And what happened? Linear or curved?
@@paulallen5321 it was of course linear. It was an academic example just to get comfortable with DoE. So i didnt expect any curved-nature in the first place. :D Thanks for your superb explanations and support!