Minitab DOE - Full Factorial Analysis

[@smigo03]

Well explained! Thanks.

[@user-lc6fg7zi5s]

Hello Paul, I would like to thank you for the interesting series that you present to the audience. I find the videos very interesting and useful. I watch them with pleasure. I have the following question related to DOE. When I create a full factorial design on two levels, for three factors, how can I plot the variance of the predicted value of the studied indicator for such a plan in a Minitab?

[@paulallen5321]

Hi there - if you watch the video I create a Standard devation column to analyse. If you want to analyse the variance instead create row statistic that works out variance and then analyse those results...

[@vivekmaripudi9174]

Nicely done. Thanks.

[@paulallen5321]

Thanks for watching!

[@dr.husseinabdelamirmohamma9642]

thanx Paul, Great job

[@paulallen5321]

Mohammad - thanks for your positive comments. You might also like my book - Design of Experiments for 21st century engineers...you can get it from LULU.COM

[@anuraggarg849]

Awesome explanation !!!!

[@paulallen5321]

Anurag - thanks for posting a comment - If want more information on my DOE approach I've just released a new book... ' Design of experiments for 21st century engineers' you can get it from LULU.COM

[@l_t_truth_seeker]

Mr. Paul ... I've got the following results ... Time= 5.02737 ... Pressure=125.081 in my Prediction! In addition to that my graphs turned out different from yours. I don't know what happened, FYI, My results for the first and second analysis are exactly the same as your results. No problems until I've got to Response Optimization. I truly appreciate you for the tutorial video and your time. My other question is: Why my "View Model" Section Default is (S)? Why isn't (Y bar)? Just like your shown settings!

[@paulallen5321]

I've also made a video about this question...load it soon..

[@l_t_truth_seeker]

@@paulallen5321 Sounds great! You made this subject fun and knowledge receiving at the same time. I like it when you say "We're all about physics not statics" ... No one mentioned that wise statement other than you.

[@paulallen5321]

Cheers - thanks, well it's too easy to concentrate on the maths, that will not solve your problem. You have to change your process and that's physics...

[@rogerb.8402]

Hi Paul, firstly, fantastic video! Just a quick question to help me explain to someone in layman terms: if optimized settings are time and pressure, could I say that the temperature can be anywhere between 70 - 90 as long as the time is 7.5 and pressure is 128.364?

[@paulallen5321]

Hi Roger - yes you could say that, so typically you'd choose the cheapest setting or use temperature to hit another target that you're also interested in...

[@rushabh695]

problem- A construction site contains 5 different facilities around the site in a particular manner and has a workforce of 25 people on site. Every person visits the facilities numerous times but not all the facilities. Total distance to travel to each facility is calculated (no.s of person X no.s of times X distance) Now if the location of those facilities on the same site is changed considering the exact same number of times the facilities are visited by the same people we calculated the new distance travelled by persons to reach the facility. Can we implement " ONE WAY ANOVA " in this case and calculated the variation of the new and old layout of the same site? ( sample size can be increased to 3 ) can this method help in optimization of construction site layout? GOALS the goal is to understand the amount of distance is being travelled by the workforce and optimize the layout of the site by locating facilities in such a manner.

[@leventnergiz]

Very Nice

[@paulallen5321]

Cheers levent - Don't forget to get my new book on DOE... Design of Experiments for 21st Century engineers...from LULU.COM...

[@BondeJakob101]

Hi Paul, when you say you want to hit a target in the video @10:10, you talk about a conformation run, where you want to hit some midpoints, then you talk about centre points: TIME 7,5 TEMP 80 PRESSURE 150, and the you show the response optimization and then you say you want to hit 95, why 95? where does the 95 come from and what is it?

[@paulallen5321]

Good to hear from you...95 would be the predicted output when I set the inputs to 7.5 80 & 150. It would whatever you're measuring as a response from the DOE. The mid point of the inputs is the weakest point of the model if the response is not linear it will show up most at this point...

[@shivsuresh1558]

Hi Paul.. this is great.. One small suggestion.. When you make a continuous video like part 1 , 2 etc.. Can you provide the link in description so that it will be easy to navigate to the previous video.. :) Thanks.

[@paulallen5321]

Hi Shiv - thanks for the comment, I have now made sure that all the minitab video's are in a play list. Not all the video's where in the playlist and this one was missing. So if you go to the channel and look for minitab they are in that playlist...cheers.

[@shivsuresh1558]

Hi @paulallen in the second video the data set is not the same.. Can you confirm when did you change the data set? because the Row 2 in your first video is : Time 10; TEMP 70; Pressure 100 & y1 : -0.05 however in this video it start with different one Time 10; TEMP 70; Pressure 100 & y1 : 164.35 .. that means did you replace the Factors as well?

[@paulallen5321]

Shiv - the reason for the transformation is the way different ways that software set up DOE's. One column is always split 50/50 hi/lo in a 2 level DOE. In my software the column is the first one in Minitab it's the last one so I had to transform the data to get the answer...

[@MissAsdfb99]

Hi, In the mac , analyze factorial design selection is not appearing directly. how can I solve this ?

[@paulallen5321]

Hi Fiami - I think you should call minitab they have a great help desk...

[@rogerb.8402]

Hi Paul, you referred to a video that you did before this video. I took a look but just to be safe, can you reply with the video that you are referring to when you set things up? Thanks! :-)

[@paulallen5321]

Hi Roger - here is the link..kzbin.info/www/bejne/q3vIqIiihpakjrc

[@rogerb.8402]

@@paulallen5321 Mucho gracias! I am referring my students to your site to help them prepare. Great work! :-)

[@paulallen5321]

Roger - Many thanks for your support...

[@olagbemioloruntegbe405]

I have a confusing question maybe I am not understanding it. Using a 3-level experimental design with 2 independent variable. what will my factor be 2^3 or 3^2

[@paulallen5321]

3^2

[@vincentmichel1827]

Hello Paul, thank you for your video. I have a question about the analysis: what is the diference between T-value and P Value ? Second point, how is calculate the effect in the pareto chart do they have a unit or a clear explication about it, please ? Thank you in advance

[@paulallen5321]

Vincent good to hear from you. tricky to answer in a quick answer. t-value realtes to where we are on the normal distribution, think z value. P is the probability of being at that point or beyond....The pareto is based on the size of the effect. So the vlaue is the same as the effect column when you do the regression...hope that helps...

[@vincentmichel1827]

@@paulallen5321 Thank you it helps

[@vincentmichel1827]

@@paulallen5321 Hello Paul, I don't find the information on minitab website. Do you know how calculate the standardized effect please ? I found something with the means and N, but don't find the right value... Please

[@salwaasyifa6277]

hello Mr. Paul. thanks for this very valuable knowledge,, please ask permission: I have a full factorial of three factors, (enzyme concentration, temperature and pH) but the general linear model results The following terms cannot be estimated and were removed: temperature; pH. if like this what is wrong with my result?. thank you Mr.

[@paulallen5321]

Salwa - please send me a mail with your results paul'[e-mail]

[@ronan437]

Hi Paul, Thanks for the videos there very helpful. I have a question around the response optimiser. When you input your target , min , and max . Minitab gives you optimum settings (factors) to hit you’re target. Is there a way of determining the max / min settings (factors) before the response is out of tolerance. Essentially determine you’re process window of low/nom/high factors before you run into trouble. Any input would be appreciated. Thanks

[@paulallen5321]

Ronan - good to hear from you. You have to remember that there is notaural variability without the input factors moving. Best thing you can do is to fix the controlled factors exactly where minitab tells you. However if you wanted to set tolerances you would have to run a monte carlo simulation. Quantum software from sigmazone can do this really well... you could do it with Minitab but it take about an hour to set up and compute...

[@ronan437]

Paul Allen Thanks for the reply Paul. So essentially response optimiser should only be used to determine you’re nominal settings.

[@paulallen5321]

Ronan - that's a good way to put it. But with process setting there should be no variability. If your DOE says the speed should be be at 253. if you dial that into the control panel why would it move?? It only moves if you allow the operator to 'twiddle'. The best companies in the world never that to happen, so it is a nominal setting, but for settings there should be no movement...

[@ronan437]

Paul Allen Understood, however I work in the plastics industry. Therefore the DOE will give me optimum settings to achieve dimensional criteria. Which is great. However we need a process window around these optimum settings to tweak the process for cosmetic issues. Without falling out of specification on the dimensions .!

[@andreamartinelli8185]

Hi Paul. I've understood everything, but I don't understand a lot of stuff that are basic concepts. For example, I don't understand, despite for the effect value, the meaning of all other values present in the tables, like T-Value, P-value excetera. Do you have any suggestions? I would like to use DOE, but I don't want to lose myself on statistics, do you have any suggestions on which things to study?

[@paulallen5321]

Andrea - If want more information on my DOE approach I've just released a new book... ' Design of experiments for 21st century engineers' you can get it from LULU.COM

[@JoaquinBS92]

Hi Paul, could you kindly explain why you run the experiment taking the mean instead of directly using the replicates?

[@paulallen5321]

Hi Joaquin - good to hear from you. That's a really good question. Normally minitab likes to put all the individual data points in a single column. I hate this method, I like to be able to look across a row of data to if there are any strange data points. That means that I have to calc row statistics, which also means my P-values are not quite right. But I prefer this when I use Minitab. If I use better software (DOE Pro) which I usually do, I don't have to use this work around...

[@JoaquinBS92]

@@paulallen5321 thank you for your quick response Paul! I agree with you that with that format it is very easy to look through the data to spot inconsistencies, thank you for that wisdom pill. My doubt came because I think with this method we are losing the within group variation, and I was wondering the effect of it... I would be interested to know your thoughts about it?

[@RaciocinioPlease]

Hi there, very clear and useful content first of all. I would like to ask you how would you define the hypothesis in an experiment like this 2x3?, is there any concrete material you could refer me to where they address this?, I am not sure I have found this in the Montgomery´s books for example, thanks a lot

[@paulallen5321]

Hi Raciocinio - do you mean what mathematical hypothesis test do you use? or how do you specify the hypothesis verbally?

[@RaciocinioPlease]

@@paulallen5321 I mean how to specify/define the hypothesis verbally...for example in my case I have 3 factors like: Wall thickness of the tubes, Diameter of the tubes and diameter of the DIE, each at 2 levels and I am analyzing their potential influence over, for example, the final roundness of the tubes after the drawing process...So I am not sure, how many H0, H1 to define in this case and how?, thanks a lot for your quick feedback, it helps a lot

[@paulallen5321]

It's not soemthing that I feel the need to do. But if you want just specify each individually as if you're only testing one of them, wall thickness v roundness for example. Because in a DOE that is exactly what happening anyway...

[@RaciocinioPlease]

@@paulallen5321 Hi, thanks for your response. I get what tou mean. I would like to ask you something else. As mentioned I am working on a design with 3 factors at 2 levels each, however I am trying to analyze how these influence not 1 but 3 response variables. Is it possible to address this like 3 separate designs because of having these 3 response variables? or is it possible just to have a design with all 3 response variables at the same time in Minitab? is it even possible?, thanks a lot for any detail on this or a material where this specific case is covered, best regards, R.

[@paulallen5321]

Hi - yes you can analyse 3 models at the same time. You'll need 3 sets of output data. If you tell minitab to analyse 3 columns of data, one for each output then when you use the optimiser you can ask it to hit 3 targets one for each output if it can do that it will recommend the settings, if not it will tell you it's not possible...let me know how it goes and I'll help more if needed...

[@rahmanfebrianto5694]

Hi Mr. Paul. I want to ask, what do we do when it has more than 2 responses? and how do we know the optimal combination of these responses. thanks

[@paulallen5321]

Rahman - great to hear from you. Usually you are using software and asking it to use the mathematical model to hit the 3 target areas. The software I use can model 30 targets at the same time. Although I would think that when you do that it is unlikely that all 30 targets can be hit all at the same time, so you'll have to compromise..

[@rushabh695]

Can we use anova for finding variance between distance of construction site facilities layout with different types of site layout models

[@paulallen5321]

Thanks for posting a comment it always helps, If want more information on my DOE approach I've just released a new book... ' Design of experiments for 21st century engineers' you can get it from LULU.COM

[@dr.shivajithorat1027]

Hi Paul, great job. I need you suggestion for lean six sigma. I am working in a Pharma company on Chemical process optimization. So kindly let me know which lean six sigma course(provider) would be relevent for me for learning as well as for certificate? TIA

[@paulallen5321]

Great to hear from you Shivaji - Where in the world are you?

[@dr.shivajithorat1027]

Hi Paul, I am in India and willing to pursue lean six sigma online.

[@paulallen5321]

Shivaji - I don't do an on line course and unfortunately I'm not a fan of any organisation who deliver a course. You can earn a certifcate at a number of places but that won't mean you're any good at any of this.... I'm also have no knowledge of the indian market and Indian consultants. I have seen Janam Sandhu she has many many video's, but I have no idea if she a top level six sigma consultant....

[@dr.shivajithorat1027]

Okay Paul, thank you very much for your replies.

[@user-yl2cf7jn7b]

How can I select the values of the center points if I want to use them do avoid the single replicate error?

[@paulallen5321]

Not sure I answered this for you, usually 3 mid points. Run 1 as the first test one in the middle and one as the last test...this helps you to see that your experiment was well controlled. If you think you'd like to support the making of these free video's take look at my donation page. buymeacoffee.com/osId2T0BL Many thanks...

[@danishahmad8673]

how to analyze when e.g one factor has 3 levels and the other two has 5 levels. Any relevant video plz share the link in reply.

[@paulallen5321]

Ahmad - Drop me a line on [e-mail] and I'll try to answer your question thoroughily.

[@danishahmad8673]

@@paulallen5321 fine sir. I just sent you an email having all the details. Thanks

[@bhaskaransundaram3287]

hi paul, what was this project name? purpose of this project?

[@paulallen5321]

Hi Bhaskaran - It wasn't a specific project, it is data from a simulator i use in class for my delegates to practice DOE. If you want the data I think I could find it for you. [e-mail]

[@bhaskaransundaram3287]

@@paulallen5321 Please Paul i need that datas to practice DOE along with you based on video instructions. my mail id - [e-mail]

[@rogerb.8402]

Hi you are confusing replicates with repeats. Not a major thing but what you are showing are repeats. A replicate would be doing the entire run again :-)

[@paulallen5321]

Roger - you are correct, I prefer not to differentiate with my client to keep it simple...and I get them control the process in such a way that there is very little differcne between the 2...

[@rogerb.8402]

@@paulallen5321 Cool, I like that!! How does minitab calculate results from replicates vs repeats? Is it the same? :-)

[@paulallen5321]

Roger - If you simply allow minitab to create a column of data then it doesn't know the DOE data was collected as repeats not replicates..so the analysis is the same...