Comparison of several criteria to decide on the significance of effects in a robustness test with an asymmetrical factorial design

A robustness test was performed for the reversed-phase HPLC method to determine the total content of triadimenol. It is based on an asymmetrical factorial design that allows examination of simultaneously a factor (column manufacturer) at four levels and 12 factors at two levels. Different error estimates applied in a t-test statistic to decide on the significance of the factor effects were compared. Dummy factor effects and experiments at the nominal level of the quantitative factors were used to derive critical effects. Moreover, methods that estimate the critical effects from the distribution of the non-significant effects were also applied. These statistical methods are compared with a graphical evaluation by means of the half-normal plot. (C)2000 Elsevier Science B.V. All rights reserved.