Greetings,
I'm doing a screening design and want to allow for lack of fit. I need replications so I want to add four additional points to my dataset. Rather than use center runs I use corner points which I am most interested in studying.
How does this affect the computation of the effects in a 2^(p-k) fractional factorial design.
The average effect is computed as the difference of the sum of the effects at a high level minus the sum of the effects at a low level. The denominator is (1/2)*(k-p)*n. Does (k-p)/2 here represent the number of pluses and minuses in a factorial effect column for one replicate and if I add 2 more plusses and 2 more minuses will the denominator change to (k-p)/2+2 where n =1 (number of replicates). I am effectively doing partial replication.
Thank you.
Mary A. Marion
The design of experiments should be under the control of balance and orthogonality, where balance indicates that the number of plus is the same as of minus, and orthogonality reflects the index for factor's interactions. In addition, center points play a statistically significant role as the standard for the experiment's pure error from its replication. Lack of fit test is for the other errors than pure ones.
Your choice did not follow these principles, which may be not computed as DoE.
- Badges
- Science topic
- Similar topics
- Engineering
- Computer Engineering
- Computer