The power of test of the treatment effects will get higher when sum of squares of degree of freedom of more than 2 is divided into each component having degree of freedom of 1 by means of orthogonal contrasts.
The factor effects obtained from experimental data such as orthogonal array experiments can be considered in detail by establishing contrasts based on the own intrinsic technology. Thus, this approach can lead to the significant conclusion in detail and practically.
This approach is better for power of test than general analysis of variance as results of the numerical simulations.
In case of saturated designs, usual analysis of variance is difficult. On the other hand, this approach is appliable because the degree of freedom of error is increasing as the sum of squares of degree of freedom of 1 increases instead of the sum of squares of degree of freedom of more than 2.
Applying this approach will be able to support applications to practical technical affairs including treatments having beyond three levels. The profit of this approach is big enough for experimenters.
View full abstract