For three-component system, regression models can be generally set in the form of polynomials which are usually defined by the following canonical Scheffe forms:
• linear model
• square model
• incomplete cube model
• complete cube model
• incomplete quartic model
• complete quartic model.
From the many models that meet the adequacy requirement, it is necessary to choose a model with a rational number of variables for ease of interpretation and practical application of the model. Selection is made by Whitcomb Score factors and selects the model that has a top rating. This paper presents a method for selecting the best model.