SELECTION OF EXPERIMENTAL DESIGNS

The range of variables can be limited by controls and thus prevent the discovery of statistically significant effects. For instance, if the column temperature was suspected to be critical and hence always controlled within a range that caused no major effects. Analysis of this data after its collection could lead to the conclusion that changes in temperature are insignificant. 

Variables can often be altered simultaneously, this makes it impossible to later determine which variable caused a given effect. This is known as semi-confounding of effects. 

Another effect is known as nonsense correlation, and is observed because inevitably we do not measure all the variables that can effect a method. If there is a latent or lurking variable that causes an effect in 

Other problems pointed out by Box et al. [20] are serially correlated errors, dynamic relations and feedback. All the above problems can be overcome by the use of properly designed statistical experiments that employ features such as randomisation, blocking and other suitable controls. 

Box et al. [20] provide a good introduction to factorial designs, the most thorough ruggedness test would involve the application of a full factorial design that tests all main effects and interaction effects. 

---

D. Examples of Some Factors and Their Levels SELECTION OF EXPERIMENTAL DESIGNS... 

---