This thesis compares analysis methods for the ordered categorical data obtained from orthogonal array type experiments and provides a guideline for selecting an appropriate one. The analysis techniques considered in this thesis are Nair's method, Extended Mantel-Haenszel method, Mean Response Model, and Taguchi's signal-to-noise ratio(SN) analysis. Performances of the above methods are compared by Monte Carlo simulation in terms of the power of detecting effective factors and of the possibility of declaring ineffective factors as effective ones. The results indicates that if one wants to reduce the possibility of declaring ineffective factors as effective ones, he may adopt Nair's or Extended Mantel-Haenszel method. However, if one wants to increase the possibility of detecting factors of small effects, he may adopt Mean Response Model with scores other than midrank scores. Since, in industrial setting, it may be more expensive to miss effective factors than to falsely declare ineffective factors as effective ones, Mean Response model may be recommended.
