How to constrain a categorical factor to a discrete numeric factor

[#1]

Hello

I am trying to generate a design with 6 factors in JMP, each of which has two levels. I am using two different buffers, and I want to use a separate buffer for each pH. When I use the 'disallowed combinations filter', and I constrain each pH to a specific buffer, I get an alert of  'optimal designer failed to converge'.

Here are the factors that I am using. Any help is appreciated!

Factor A (continuous)- 50, 80

Factor B (continuous)- 100, 40

Factor C (continuous)- 2, 8

Factor E (continuous)- 800, 500

Buffer (categorical)- K ,N

pH (discrete numeric)- 11, 7.5

I want to use buffer K for the pH at 7.5 and buffer N for the pH at 11.

[#2 · Accepted solution]

Hi @LambdaAvgBee104,

If buffer and pH are closely related and can not vary independently, why would you add these two factors in the design ? It seems the information for the design comes only from one factor.

Looking at your design and study, it seems that you only want to use buffer K at pH 7,5 and buffer N at pH 11, so it doesn't seem possible to discriminate the effects of pH or buffer on the responses, as you may need different pH values to assess the effect of pH on the responses independently from the buffer chosen.

You could maybe remove the pH factor and only use the buffer categorical factor which will contain two experimental informations : the buffer type used and the resulting pH value in which the specific buffer is used.

Does that make sense for you or did I miss something ?

PS: The snapshot you give does not correspond to the situation you described, as you need to specify what are the combinations NOT to be included in the design. In your case, you'll have something like : 

Note that I don't generally put the limit exactly on the threshold values (7,5 and 11 in this example) in order for the exchange coordinate alogrithm in the Custom Design to be able to converge (but here this is not the only problem, as factors buffer and pH are completely confounded). 

[#3 · Accepted solution]

I agree with Victor.  You have a 5 factor experiment.  One of the factors is pH/Buffer.  It has 2 levels: 7.5K and 11N.