Abstract
The four experimental designs were investigated as typical examples of the non-orthogonal experimental designs. These four cases reduced the number of experiments than the applicable orthogonal design. The experimental efficiency of each design was only slightly smaller than the orthogonal design.
Four cases, that is, the designs in which several experiments were omitted from the assumed orthogonal arrays intentionally, or in which several experiments were added or expanded were also investigated by numerical simulations. As results, it could be confirmed that there are no big problems with the powers of test of non-orthogonal experimental designs practically.
As there are no serious bottlenecks in the practical applications of non-orthogonal experimental design adopted to engineering affairs, the profit by reducing the number of experiments is fairly large.
When experimenters could not assign factors to the assumed orthogonal array because of lack of columns, they will adopt one more bigger orthogonal array. Minimization of the number of experiments will agree with the request of experimenters, and has much worth.
Though the precision of estimates depends on adopted non-orthogonal experimental design, non-orthogonal design can make it equal to the corresponding orthogonal design about the specific important factor.
