The Design for Trustworthy Software Compilation Taguchi Methods and Optimization for Robust Software

Orthogonal Matrices for Taguchi Parameter Design Experiments

This section presents more technical details for the use of the reduced factorial or orthogonal matrix method so central to Taguchi Methods. The full nomenclature for an orthogonal array is L_a(b^c), in which L stands for Latin square, a is the number of test trials, b is the number of levels for each column, and c is the number of columns in the array. An orthogonal matrix is a generalized Latin square. For example, the L₄(2³) orthogonal array is shown in Table 7.

Table 7 An L₄(2³) Orthogonal Array

Appendix B of the Robust Engineering Workshop Manual from the American Supplier Institute ...