How F and P values are influenced by centring

Chemometricians are often much concerned by column transformations, especially centring. There is relatively little literature on how centring can effect the apparent p values of a model. Most packages, such as Excel, allow choices whether to centre data and also to include intercept terms, but it is important to understand how this influences decisions as to whether a model is significant or not. In previous articles, we have always chosen designs where the x block is centred. In practical terms, regression models are often chosen where neither the x nor y block is centred. Table 1 illustrates a simple dataset.