Can I add a factor to an existting DOE and/or change the values of any run in an existting DOE?

[#1]

Hi, I developed a DOE model with 3 factors (one categorical) for a specific time (7 days) containing 16 runs. Later I used the same DOE but now for samples (runs) with a loger time (28 days). Now, I have decided to add the time as a factor but I want to keep the runs that a I have (16 runs for 7 days, and 16 runs for 28 days); to improve the quality of the model I decided to add center points to the design (time 17.5 days): I added 4 runs, one for each value of the categorical factor. How can I do that, keeping some previous runs? In the augmented design it is not allowed to add a new factor, or is it allowed (if yes, I do not know how to do it). So, I decided to develop a new DOE, now with 4 factors (including time as a factor) and with 30 runs (16 for 7 daus, 4 for 17.5 days, 16 for 28 days), but the table generated by jmp has some runs different from the previous 16 runs (for 7 and 28 days); can I change the values of any factor in this new DOE in order to keep the previous runs, is this alllowed wtihout compromising the accuracy? Or simply I cannot perform this analysis?

[#2]

While I'm not sure of your exact workflow and if in fact you used the augment design platform to generate the second design, fundamentally you can manually change any levels in the JMP data table that contains your design...but here's what I would do to assess your idea of '...compromising the accuracy?' Whatever that means. If in fact you used the augment design platform and it produced a resultant design, keep that design 'as is'. Then manually change the treatment combinations in a second separate design as you desire. Then use the Evaluate Design platform to see how the two designs compare from a variety of perspectives such as power, correlation of parameter estimates, etc.

[#3]

Hi, thank you for your reply. Regarding the accuracy issue, I just want to say that I do not know if any change in the JMP data table has any impact in the model's results. I am not an expert in this field, I am a beginner using jmp.

Regarding your second comment, in the augmented design the time was not an initial factor, so, it is not possible (as far as I know) to add treatment combinations (runs) because I cannot define values for time.

As you may see in the file '1-DOE (7 or 28 days)' attached, it contains 3 factors and 1 response (compressive strength, fc). This DOE conatins 16 runs and is valid either for samples cured during 7 days or cured during 28 days (at the end of curing time, 7 or 28 days, unconfined compressive strength tests were performed to characterise the fc). My question is that I want to add a new factor to this design, time, but I want also to keep the previous treatment combinations (16 for 7 days and 16 for 28 days); I just need to add center points for time, 17.5 days. Can you hlep me how to do it? Augmented design, but how to add time as a new factor keeping the previous treatment combinations (16 for 7 days and 16 for 28 days)? Or I should generate a new DOE with 30 treatment combinations and then manually change any levels in the JMP data table that contains this new design?

[#4]

I'm a bit confused.  There is very little context provided, so virtually impossible to provide advice. If I understand what you have written, you ran an experiment with 3 factors in 16 treatments (not sure the number of levels for your factors or how you got 16 treatments) and that experiment took 7 days to run?  Is that to evaluate the experimental units?  Is this because it takes a long time to make the experimental units (run a treatment) or it takes this long to measure the EU's?  You want to replicate the design over three time periods, why?  Are you interested in the change in the responses over time?  It doesn't sound like you are adding center points as this would require center points for all factors? Sounds more like you are creating 3-levels for the factor time. Why would you care about a quadratic term for time?  Or are you looking for the "best time"?

As Pete writes, what do you mean by accuracy?  

[#5]

Thank you for your comment. Please see my reply to P_Bartell comment. I hope it may carify what I am trying to do.

[#6]

Not really, there are a number of question I posed that are not answered.  You are NOT creating center points.  You are replicating the experiment over 2 (perhaps 3) cure times.  Cure time can be added to the model (factorially as well as the 2nd order polynomial).  Recognize that cure time is confounded with block effects.  I assume you are interested in how large the effect of cure time is in relation to the other factors in your design?  Or are you just trying to pick a winner?

[#7]

Thanks again for your comments. Inded curing time is a new factor to be added because I know that response (compressive strength) chenges with curing time, and the change is not linear is a second order type. Regarding center points, I want to add them for all continuous factors (2 previous factors [dust, binder_sand_ratio] + new factor curing time), one center point for each of the categorical factor (finder_type). So, I want to keep all experiments that I have performed for 7 days and also for 28 days, but I want to do a new design considering curing time as a new factor. For me, it seems that it is enough to add to the previous experiments (16 experiments at 7 days that I have + 16 experiments at 28 days that I have) 4 new experiments, center points; in total I will have a design with 30 experiments and 4 factors. How can I do this keeping the initial experiments, the initial design that was developed for 7 and 28 days? Any help is welcome.

[#8]

I don't understand your math.  "For me, it seems that it is enough to add to the previous experiments (16 experiments at 7 days that I have + 16 experiments at 28 days that I have) 4 new experiments, center points; in total I will have a design with 30 experiments and 4 factors"

16 + 16 + 4 is not 30?

You want to replicate the original design only use cure time of 28 days (just add a column called cure time and put 7 in the first 16 and 28 in the second 16.  Now center points is a bit of a challenge for all factors as one factor is a 4-level categorical factor (there is no center point).  So while you could run center points for the other 3 factors and cure time, you would have to decide what to do with the categorical factor for those center point runs.

I have attached a JMP table with 36 runs for your evaluation.

[#9]

Hi Statman, I do not know where 30 come from! Sorry for that mistake. Yes, your table seems ok, is this that I want. But my question is, remember that I am a beginer in JMP, can I do that? The jmp table should not come from the DOE design, defining first the factors, their levels, and identifying the response? Then, DOE will generate a designing table will all eperiments taking into account all factors and levels, and when I did that the jmp table was diferent from the one that I want (is different from the one that you sent to me). But reading your comment I realize that everythng seems to be easier with jmp. I see now that with this table, if I click on the 'Model' > 'Model Specification' I can identify the factors ('Construct model effects') where I can add 'curing_time' as a new factor so, my problem seems to be solved. Thank you for your precious help.