I would like to optimize my extraction using responce surface methodology but I‘m perplexed in choosing what design should I use? I just have 2 factors with 2 levels each, so I considered using CCD but I don’t understand the basic of the run test results which have some replication or about the lack of fits (is this crucial?). If I change it to no replication or minimizing lack of fits, is it ok? Or should I consider another design?
thank you very much for your kind response
Lack of fit F test is used to test the hypothesis of model correctness: that is whether there is statistically significant model misspecification in not.
In RSM, the Lack of fit test is used to determine whether there is significant curvature in the experimental data: the first order model with interaction term is not able to capture adequate data variation and hence a second order model based design (CCD or Box Behnken type) is needed to fit the data coming from this particular region of the experimental space.
Details about the RSM process and statistical tests associated with it can be found in the following books:
1) Chemometrics: Experimental Design by Ed Morgan
2) Response Surface Methodology by Myers, Montogomery and Cook
Dear Fatimah Azzahra,
If you have just 2 factors with 2 levels, I recommend you the one-factor-at-a-time (OFAT) method. In my opinion, this method is more accurate than the Response Surface Methodology for your experimental design.