Coding of non-equally spaced variables for Response Surface Model

Question

Hi there,

I am using a Response Surface Design with 3 levels for each factor. The factor levels are not equally spaced because I chose the values of the factors at the centre point so that the response is  approx 50% of the range that I am exploring. I have two doubts:

1) Choosing the factor values for centre point  to give 50% of the response range is the right way?

2) Is there any guidelines to code the factors when they are not equally spaced?

Thank you in advance

louv · Answer

It could be my confusion but I like to think of the Response Surface Design as a two-level design with center points which is what I believe you mean when you say three levels (-1, 0,+1). The values of the factors at the center point should be at the mean value of the range of factors.

ajanda · Answer

Dear LouV,

Thank you for your response. I have been shortly exploring the use of discrete numeric factors. Nevertheless, in the case that I am analysing my factors are continuous, so I do not have the limitation of discrete values.

I might posed my question in the wrong way. My question is:

the values of the factors at centre point of CDD design should be at the mean value of the range of factors?

or the values of the factors at the centre point should give a  50% of variation of response respect to the range that it is being explored?

Thank you again.

louv · Answer

Have you investigated the use of discrete numeric factors in the custom design platform. Many times we deal with factors that are numeric however they are not easily set at specific values due to operational considerations. We have the capability in the custom design platform to have a "discete numeric" choice so that such a scenario can be dealt with successfully.