Biweekly Report 16-30/7/2012
Considerable amount of time was devoted to scidoe_bbdesign.sci, the function producing a Box - Behnken design of experiments.
After making some changes to the code, the factorial points match each number of variables as stated in Wikipedia and in R, another open source statistics program.
However, the cases for k=9,16 still remain unresolved, while no known reference for these results has been found.
Also, I wasn't able to find a relation producing a blocked design, using only the -1's and 1's of the design, and not the responses of an experiment.
Scidoe_ccdesign.sci is the function for a Central Composite Design of experiments.
Two issues remain unresolved :
- The value of alpha for the orthogonal design isn't correct, so for the moment I set the default design to 'rotatable'. An important note is that a Central Composite Faced Design cannot be 'rotatable'.
- The values of the factorial cube points in the case of 'inscribed' design have to decrease, as the number of factors increases. I still haven't got the relation to correctly decrease the values.