Visual Six Sigma: Making Data Analysis Lean

6.6. Modeling Relationships

In initial brainstorming sessions to identify causes of bad parts, the team identified five possible Hot Xs:

• Coating variables: Anodize Temperature, Anodize Time, and Acid Concentration.

• Dye variables: Dye Concentration and Dye pH.

It is time to determine whether these are truly Hot Xs, and to model the relationships that link Thickness, L*, a*, and b* and these Xs. So, Sean will guide the team in conducting a designed experiment. These five process factors may or may not exert a causal influence. The designed experiment will indicate whether they do and, if so, will allow the team to model the Ys as functions of the Xs. The team will use these models to optimize the settings of the input variables in order to maximize yield.

Although Color Rating is the variable that ultimately defines yield, Sean knows that using a nominal response in a designed experiment is problematic—a large number of trials would be required in order to establish statistical significance. Fortunately, the team has learned that there are strong relationships between Thickness, L*, a*, and b, and the levels of Color Rating. Accordingly, Sean and his team decide that the four responses will be L*, a*, b*, and Thickness. There will be five factors: Anodize Temp, Anodize Time, Acid Conc, Dye Conc, and Dye pH.

With the data from this experiment in hand, Sean plans to move to the Model Relationships phase of the Visual Six Sigma Data Analysis Process (Exhibit 3.29). He will use the guidance ...