As I have 4 factors in an experiment with three levels I need to determine the levels by myself not by alpha method in central composition method. Forexample I need to have x=0.2,0.25,0.33 or t=1,12,24 and etc. Bit when I use design expert it gives me its random numbers like x=0.2,0.265,0.33. Note that experimentally I have to choose x=0.,0.25,0.33. Please let me know if there is any unclear point
They are not random numbers! The centre point is suppose to be in the middle between the low and high. So, 0.2, 0.265 and 0.33.
If the factors are parametric you really should use the default alpha. It is important to make the design rotatable so that the experiment has the same power in estimating the coefficients for the different factors.
You say that you cannot do level 0.265, why? Is it not a parameter that you can set to any value? If not, it is a categorical value. Then use coded levels -1, 0 and +1.
Or is it that you have already made the experiments with the incorrect level?
first thanks for constructive comment. I was wondering if there is any possibility to define the exact levels manually. However it seems impossible. I think you are right I have to introduce x as categoric factor but the number of experiments increases rather than defining that as numeric factor. Experimentally, x is highly recommended to be 0.2,0.25 and 0.33, and 0.265 seems weird to be practical. I didnt still start my experiment and Ill start while i could do a workable design of experiment
Why would the number of runs in the design be larger if you use a categorical variable? You can use the same design, just change that factor to be categorical and code the levels so that instead of inputting 0.2, 0.25 and 0.33, you just set them to -1, 0 +1 or "low, "centre" and "high".
you can work with coding and declare the lowest level to -1, the mid level to 0 and the highest level to +1. The calculated values for the optimum then need to be recoded.
- Badges
- Science topic