CCF blocking question

[KarenK]

Good evening,

When I set up a 3 factor CCF design with , the system recommends 12 runs (2^3 factorial plus 4 center points) for block 1 and 8 runs (6 axial points + 2 center points) for block 2, for a total of 20 runs.

We have a setup with 10 experimental units. It would be preferable to run as 2 blocks of 10. Can I move 2 center points from the 12 run set to the 8 run set and still retain the same analysis (so block 1 has 2 center points and block 2 has 4 center points), or is there some statistical basis for avoiding this? If we do so, what analysis can I do to compensate?

Thanks,

Karen

[Wayne]

When there are special considerations, such as fixed block size, constraints, high-order models, etc. the canned CCD design are not the best tool. It is probably better to use the more flexible Optimal designs to accomplish these goals. Most of the time optimal designs will have equally sized blocks.

If you choose to manually correct the CCD as specified, just check the evaluation node for aliasing and correlation, and the Graphs for an unacceptable fraction of design space score.