Courant Institute of Mathematical Sciences
Department of Computer Science
New York University
To generate a no pivot set, a typical control file looks like this:
illumination dark light sucrose 0 30 60 90 nitrate 0 5 10 15 time 1 2 4 8 16 organ root leafwhich says that the input attribute illumination has two values (dark and light), sucrose has four values etc. Invoking this by typing
k combgeneric combgeninwill produce, in the file tmpexpercsv, the following no pivot two factor design:
num,illumination,sucrose,nitrate,time,organ 1,light,90,10,16,root 2,light,0,15,16,leaf 3,dark,30,5,16,leaf 4,light,60,10,8,root 5,dark,60,0,4,leaf 6,light,30,0,1,root 7,dark,90,0,8,leaf 8,dark,0,10,1,root 9,dark,30,10,2,root 10,dark,90,15,2,root 11,dark,0,0,2,leaf 12,X,60,15,1,leaf 13,light,30,10,4,leaf 14,light,60,5,2,root 15,X,90,15,4,root 16,X,0,5,4,X 17,X,90,5,1,X 18,X,30,15,8,X 19,X,0,5,8,X 20,X,60,0,16,Xwhere the X represent don't cares.
If a pivot, say on illumination is desired, then the input looks like this:
illumination dark light sucrose 0 30 60 90 nitrate 0 5 10 15 time 1 2 4 8 16 organ root leaf pivot: illuminationInvoking in the same way as above yields:
num,*illumination*,sucrose,nitrate,time,organ 1,dark,60,15,4,leaf 2,dark,0,15,8,root 3,dark,90,0,4,root 4,dark,30,5,4,leaf 5,dark,60,0,2,root 6,dark,90,5,8,leaf 7,dark,30,0,8,root 8,dark,0,0,1,leaf 9,dark,90,15,1,root 10,dark,0,10,4,root 11,dark,0,5,2,root 12,dark,0,5,16,root 13,dark,30,15,16,leaf 14,dark,30,10,2,leaf 15,dark,60,5,1,leaf 16,dark,90,15,2,leaf 17,dark,60,0,16,root 18,dark,90,10,16,root 19,dark,60,10,8,root 20,dark,30,10,1,root 21,light,60,15,4,leaf 22,light,0,15,8,root 23,light,90,0,4,root 24,light,30,5,4,leaf 25,light,60,0,2,root 26,light,90,5,8,leaf 27,light,30,0,8,root 28,light,0,0,1,leaf 29,light,90,15,1,root 30,light,0,10,4,root 31,light,0,5,2,root 32,light,0,5,16,root 33,light,30,15,16,leaf 34,light,30,10,2,leaf 35,light,60,5,1,leaf 36,light,90,15,2,leaf 37,light,60,0,16,root 38,light,90,10,16,root 39,light,60,10,8,root 40,light,30,10,1,root
The combinatorial design software runs in a high performance interpreted environment called K.
Our papers give several references of similar work such as Taguchi designs. The direct inspiration comes from the AETG work on software testing from Telcordia. Here is a link.
This work was partly supported by the U.S. National Institutes of Health (Grant no. GM32877) by the U.S. National Science Foundation (Grant no. IIS-9988636, MCB-0209754) and a U.S.National Research Service Reward (grant no. GM65690). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the supporting organizations. This support is greatly appreciated.