Extend a DOE with new discrete numeric values

[#1]

Hi,

I have performed a DoE on a chemical reaction with four parameters, three with four discrete numeric values and one constant. The DoE gave me some useful insight and I now want to extend the DoE but limit the values of my discrete parameters and change my constant value to a continuous. I've watched many of your YouTube videos and searched the discussions here without luck.

I found that my lower value for one of the parameters was the best (let's call it A out of A, B, C, and D), so now I want to extend the DoE with a more narrow range around A and remove B, C, and D from the upcoming experiments. When trying to using augment design I am not allowed to change the factor to A, E, and F. In line with this, I am also not allowed to change my constant parameter to a continuous for the following experiments.

I must have missed something, but I feel like I've looked everywhere. Is the only solution to make a new DoE with my newly desired parameters or can I extend my current DoE with these?

[#2]

I caution you against narrowing the factor ranges. Why? The DOE generates the best data to estimate model parameters. You use the model, in turn, to estimate the best conditions. Limiting the range will compromise the precision and power of the parameter estimates.

The factor characteristics are saved as column properties. Change the property settings before augmenting the design.

[#3]

I don't understand your situation and you have provided little detail from your first experiment, but here are my thoughts/questions.  First why do you care to "extend" the first experiment?  It seems like factor A is the only factor you want to include in the next experiment.  The inference space may be completely different.  

1. For your factor "A" you found that the lower value was best.  How many levels was this factor tested at? It seems like you want to experiment with an even lower value of A?  Were there no interactions?

2. You discovered B, C and D were not useful (e.g., significant), so you want to drop them from consideration.  How "bold" were the other factor levels in your experiment? Where will you set them?

3. You want to add 2 new factors E & F and then also add the factor that was constant win the first experiment.  While this can be done, you will miss any possible interaction with the B, C and D factors.  Also the significance of A is conditional.  Its significance depends on the other terms in the model and how you estimated the MSE.  If either of those change so may the importance of A.

4. In retrospect, you should have experimented on the constant , E and F in addition to A, B, C, D in the first experiment.  Then decisions to remove terms are more effective.