Influence of the sequence of experiments in a ruggedness test when drift occurs

When drift occurs, the sequence of the experiments of a design can influence the effects estimated for the different factors examined. One can expect that the ageing of a Chromatographie column can cause a drift in some responses. Different design sequences were performed in order to examine the influence of the drift on the estimated effects of the design factors and on the interaction effects, since the latter are used to estimate the experimental error for the statistical interpretation of the effects. It was observed that several effects estimated from a design could be confounded with the drift. The extent to which effects are influenced in practice depends on the sequence in which the experiments of a design are executed, as well as on the ratio between the magnitude of the drift and the effect of the design factors. Anti-drift designs have a drawback that interaction effects are confounded with the drift. This leads to an overestimation of the critical effects. Correction of the measured results, based on repeated measurements at nominal levels, yields effects that are similar for the different designs regardless of the sequence in which the experiments of the design were performed. The drawback of this approach is that more experiments are required.