14.03.2014 Views

Modeling and Multivariate Methods - SAS

Modeling and Multivariate Methods - SAS

Modeling and Multivariate Methods - SAS

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

238 Analyzing Screening Designs Chapter 8<br />

Statistical Details<br />

Note that the three active factors have been highlighted. One other factor, X18, has also been highlighted. It<br />

shows in the Half Normal plot close to the blue line, indicating that it is close to the 0.1 cutoff significance<br />

value. The 0.1 critical value is generous in its selection of factors so you don’t miss those that are possibly<br />

active.<br />

The contrasts of 5.1, –3, <strong>and</strong> 1.8 are close to their simulated values (5, –3, 2). However, the similarity of<br />

these values can be increased by using a regression model, without the effect of orthogonalization.<br />

The p-values, while useful, are not entirely valid statistically, since they are based on a simulation that<br />

assumes orthogonal designs, which is not the case for supersaturated designs.<br />

Statistical Details<br />

Operation<br />

The Screening platform has a carefully defined order of operations.<br />

• First, the main effect terms enter according to the absolute size of their contrast. All effects are<br />

orthogonalized to the effects preceding them in the model. The method assures that their order is the<br />

same as it would be in a forward stepwise regression. Ordering by main effects also helps in selecting<br />

preferred aliased terms later in the process.<br />

• After main effects, all second-order interactions are brought in, followed by third-order interactions, <strong>and</strong><br />

so on. The second-order interactions cross with all earlier terms before bringing in a new term. For<br />

example, with size-ordered main effects A, B, C, <strong>and</strong> D, B*C enters before A*D. If a factor has more than<br />

two levels, square <strong>and</strong> higher-order polynomial terms are also considered.<br />

• An effect that is an exact alias for an effect already in the model shows in the alias column. Effects that<br />

are a linear combination of several previous effects are not displayed. If there is partial aliasing ( a lack of<br />

orthogonality) the effects involved are marked with an asterisk.<br />

• The process continues until n effects are obtained, where n is the number of rows in the data table, thus<br />

fully saturating the model. If complete saturation is not possible with the factors, JMP generates r<strong>and</strong>om<br />

orthogonalized effects to absorb the rest of the variation. They are labeled Null n where n is a number.<br />

For example, this situation occurs if there are exact replicate rows in the design.<br />

Screening as an Orthogonal Rotation<br />

Mathematically, the Screening platform takes the n values in the response vector <strong>and</strong> rotates them into n<br />

new values. The rotated values are then mapped by the space of the factors <strong>and</strong> their interactions.<br />

Contrasts = T’ × Responses<br />

where T is an orthonormalized set of values starting with the intercept, main effects of factors, two-way<br />

interactions, three-way interactions, <strong>and</strong> so on, until n values have been obtained. Since the first column of<br />

T is an intercept, <strong>and</strong> all the other columns are orthogonal to it, these other columns are all contrasts, that<br />

is, they sum to zero. Since T is orthogonal, it can serve as X in a linear model. It does not need inversion,<br />

since T’ is also T -1 <strong>and</strong> (T’T)T’. The contrasts are the parameters estimated in a linear model.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!