Abstract
This report discussed with the practical method of non-orthogonal optimal design and its power of test to promote and to support technical applications.
The main effects of two levels of factors A, B, C and D have been already known and the attention is focused on interactions of A×B and C×D.
In case of conducting this design by a usual orthogonal array, it will be experiments of L16.
Because of restrictions such as material's preparation, a due date, a cost and efforts, etc., 16 experiments would not be permitted but about 8 experiments will be allowed to an experimenter, the non-orthogonal optimal design which meets this situation most is desirable.
When choosing 8 experiments from 16 experiments of all combinations of 4 factors, there are 12,780 cases.
The allocation in which the interactions of A×B and C×D noticed above are orthogonal each other should be chosen and adopted.
As the results of simulation about the power of test in non-orthogonal optimal designs, using this non-orthogonal optimal design seemed to have no problems for the practical usage.
