Screening designs effects interpretation

Full factorial designs at two levels are mainly used for screening, that is, to determine the influence of a number of effects on a response, and to eliminate those that are not significant, the next stage being to undertake a more detailed study. Sometimes, where detailed predictions are not required, the information from factorial designs is adequate, at least in situations where the aim is fairly qualitative (e.g. to improve the yield of a reaction rather than obtain a highly accurate rate dependence that is then interpreted in fundamental molecular terms). 

P,5 is thus confounded with two other first-order interaction effects, as well as a number of higher order interactions. In addition the estimators for the main effects show that these also are each confounded with a number of interactions of different degrees. In practice it would probably be assumed that all second and higher order interactions could be neglected. This still leaves each main effect confounded with 2 first-order interactions, and this fact must be taken into account in the interpretation of the results. This may mean that certain estimates of the main effects will be viewed with some caution. The above design is essentially used for screening, as described in chapter 2. 

---